Geotechnical Aspects of Underground Construction in Soft Ground

GEOTECHNICAL ASPECTS OF UNDERGROUND CONSTRUCTION IN SOFT GROUND
PROCEEDINGS OF THE 6TH INTERNATIONAL SYMPOSIUM (IS-SHANGHAI 2008),

SHANGHAI, CHINA, 1012 APRIL 2008
Geotechnical Aspects of
Underground Construction
in Soft Ground
Editors
C.W.W. Ng
Hong Kong University of Science and Technology,
Hong Kong Special Administrative Region
H.W. Huang & G.B. Liu

Tongji University, Shanghai, China
CRC Press/Balkema is an imprint of the Taylor & Francis Group, an informa business
2009 Taylor & Francis Group, London, UK
Typeset by Charon Tec Ltd (A Macmillan Company), Chennai, India
Printed and bound in Great Britain by Cromwell Press Ltd, Trowbridge, Wiltshire.
All rights reserved. No part of this publication or the information contained herein may be reproduced, stored
in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, by photocopying,
recording or otherwise, without written prior permission from the publishers.
Although all care is taken to ensure integrity and the quality of this publication and the information herein, no
responsibility is assumed by the publishers nor the author for any damage to the property or persons as a result
of operation or use of this publication and/or the information contained herein.
Published by:
CRC Press/Balkema
P.O. Box 447, 2300 AK Leiden, The Netherlands
e-mail: Pub.NL@taylorandfrancis.com
www.crcpress.com www.taylorandfrancis.co.uk www.balkema.nl
ISBN: 978-0-415-48475-6 (Hardback)

ISBN: 978-0-203-87998-6 (eBook)
Geotechnical Aspects of Underground Construction in Soft Ground Ng, Huang & Liu (eds)
2009 Taylor & Francis Group, London, ISBN 978-0-415-48475-6
Table of Contents
Preface
XIII
Sponsors
XV
Special lectures
Processes around a TBM
A. Bezuijen & A.M. Talmon
Supporting excavations in clay from analysis to decision-making

M.D. Bolton, S.Y. Lam & A.S. Osman
15
Overview of Shanghai Yangtze River Tunnel Project

R. Huang
29
Underground construction in decomposed residual soils

I.M. Lee & G.C. Cho
45
General reports
Safety issues, risk analysis, hazard management and control
C.T. Chin & H.C. Chao
67
Calculation and design methods, and predictive tools

F. Emeriault & R. Kastner
77
Analysis and numerical modeling of deep excavations

R.J. Finno
87
Construction method, ground treatment, and conditioning for tunneling

T. Hashimoto, B. Ye & G.L. Ye
99
Physical and numerical modelling

P.L.R. Pang
109
Case histories
A. Sfriso
121
Theme 1: Analysis and numerical modeling of deep excavations

Optimization design of composite soil-nailing in loess excavation
G.M. Chang
133
Three-dimensional finite element analysis of diaphragm walls for top-down construction

J. Hsi, H. Zhang & T. Kokubun
141
Numerical evaluation of dewatering effect on deep excavation in soft clay

L. Li & M. Yang
147
Analysis of the factors influencing foundation pit deformations

Y.Q. Li, K.H. Xie, J. Zhou & X.L. Kong
153
Construction monitoring and numerical simulation of an excavation with SMW

retaining structure
Z.H. Li & H.W. Huang
159
A simplified spatial methodology of earth pressure against retaining piles of pile-row

retaining structure
Y.L. Lin & X.X. Li
165
Consideration of design method for braced excavation based on monitoring results

H. Ota, H. Ito, T. Yanagawa, A. Hashimoto, T. Hashimoto & T. Konda
173
Ground movements in station excavations of Bangkok first MRT

N. Phienwej
181
Numerical modelling and experimental measurements for a retaining wall of

a deep excavation in Bucharest, Romania
H. Popa, A. Marcu & L. Batali
187
3D finite element analysis of a deep excavation and comparison with in situ measurements
H.F. Schweiger, F. Scharinger & R. Lftenegger
193
The effect of deep excavation on surrounding ground and nearby structures

A. Sieminska-Lewandowska & M. Mitew-Czajewska
201
Multi-criteria procedure for the back-analysis of multi-supported retaining walls

J. Zghondi, F. Emeriault & R. Kastner
207
Monitoring and modelling of riverside large deep excavation-induced ground movements in clays
D.M. Zhang, H.W. Huang & W.Y. Bao
215
GPS height application and gross error detection in foundation pit monitoring
H. Zhang, S.F. Xu & T.D. Lu
223
Study on deformation laws under the construction of semi-reverse method

J. Zhang, G.B. Liu & T. Liu
227
Comparison of theory and test on excavation causing the variation of soilmass strength
J. Zhou, J.Q. Wang & L. Cong
235
Theme 2: Construction method, ground treatment, and conditioning for tunnelling

Ten years of bored tunnels in The Netherlands: Part I, geotechnical issues
K.J. Bakker & A. Bezuijen
243
Ten years of bored tunnels in The Netherlands: Part II, structural issues
K.J. Bakker & A. Bezuijen
249
The influence of flow around a TBM machine

A. Bezuijen & K.J. Bakker
255
Mechanisms that determine between fracture and compaction grouting in sand

A. Bezuijen, A.F. van Tol & M.P.M. Sanders
261
Research of non-motor vehicle-rail transit-tube interchanging transport system pattern

A.Z.G. Deng & Q.H. Zhang
269
Shotcrete excavations for the Munich subway Comparison of different methods

of face support in settlement sensitive areas
J. Fillibeck & N. Vogt
Fracturing of sand in compensation grouting
K. Gafar, K. Soga, A. Bezuijen, M.P.M. Sanders & A.F. van Tol
VI
275
281
Historical cases and use of horizontal jet grouting solutions with 360 distribution and
frontal septum to consolidate very weak and saturated soils
G. Guatteri, A. Koshima, R. Lopes, A. Ravaglia & M.R. Pieroni
287
The effects of sample dimension and gradation on shear strength parameters of

conditioned soils in EPBM
M. Hajialilue-Bonab, M. Ahmadi-adli, H. Sabetamal & H. Katebi
295
Experimental study on compressibility behavior of foamed sandy soil

M. Hajialilue-Bonab, H. Sabetamal, H. Katebi & M. Ahmadi-adli
Study on earth pressure acting upon shield tunnel lining in clayey and sandy grounds based
on field monitoring
T. Hashimoto, G.L. Ye, J. Nagaya, T. Konda & X.F. Ma
301
307
The double-o-tube shield tunnel in Shanghai soil

C. He, L. Teng & J.Y. Yan
313
Frozen soil properties for cross passage construction in Shanghai Yangtze River Tunnel
X.D. Hu & A.R. Pi
319
The influence of engineering-geological conditions on construction of the

radioactive waste dump
J. Kuzma & L. Hrustinec
325
Critical ventilation velocity in large cross-section road tunnel fire

Z.X. Li, X. Han & K.S. Wang
331
Metro tunnels in Buenos Aires: Design and construction procedures 19982007

A.O. Sfriso
335
Study on the earth pressure distribution of excavation chamber in EPB tunneling

T.T. Song & S.H. Zhou
343
Backfill grouting research at Groene Hart Tunnel

A.M. Talmon & A. Bezuijen
349
Longitudinal tube bending due to grout pressures

A.M. Talmon, A. Bezuijen & F.J.M. Hoefsloot
357
Theme 3: Case histories

Tunnel face stability and settlement control using earth pressure balance shield
in cohesionless soil
A. Antiga & M. Chiorboli
365
Displacements and stresses induced by a tunnel excavation: Case of Bois de Peu (France)
S. Eclaircy-Caudron, D. Dias & R. Kastner
373
Shield tunneling beneath existing railway line in soft ground

Q.M. Gong & S.H. Zhou
381
Case history on a railway tunnel in soft rock (Morocco)

A. Guiloux, H. Le Bissonnais, J. Marlinge, H. Thiebault, J. Ryckaert, G. Viel,
F. Lanquette, A. Erridaoui & M.Q.S. Hu
385
Observed behaviours of deep excavations in sand

B.C.B. Hsiung & H.Y. Chuay
393
Environmental problems of groundwater around the longest expressway tunnel in Korea

S.M. Kim, H.Y. Yang & S.G. Yoon
399
VII
Measurements of ground deformations behind braced excavations

T. Konda, H. Ota, T. Yanagawa & A. Hashimoto
405
Research on the effect of buried channels to the differential settlement of building

D.P. Liu, R. Wang & G.B. Liu
413
Performance of a deep excavation in soft clay

G.B. Liu, J. Jiang & C.W.W. Ng
419
Deformation monitoring during construction of subway tunnels in soft ground

S.T. Liu & Z.W. Wang
427
The construction and field monitoring of a deep excavation in soft soils

T. Liu, G.B. Liu & C.W.W. Ng
433
Excavation entirely on subway tunnels in the central area of the Peoples Square
Y.B. Mei, X.H. Jiang, Y.M. Zhu & H.C. Qiao
441
The benefits of hybrid ground treatment in significantly reducing wall movement:

A Singapore case history
N.H. Osborne, C.C. Ng & C.K. Cheah
447
3D deformation monitoring of subway tunnel

D.W. Qiu, K.Q. Zhou, Y.H. Ding, Q.H. Liang & S.L. Yang
455
Challenging urban tunnelling projects in soft soil conditions

H. Quick, J. Michael, S. Meissner & U. Arslan
459
Supervision and protection of Shanghai Mass Rapid Line 4 shield tunneling across
the adjacent operating metro line
R.L. Wang, Y.M. Cai & J.H. Liu
Kowloon Southern Link TBM crossing over MTR Tsuen Wan Line tunnels in HKSAR
K.K.W. Wong, N.W.H. Ng, L.P.P. Leung & Y. Chan
Application of pile underpinning technology on shield machine crossing through pile
foundations of road bridge
Q.W. Xu, X.F. Ma & Z.Z. Ma
465
471
477
Characteristics of tunneling-induced ground settlement in groundwater drawdown environment

C. Yoo, S.B. Kim & Y.J. Lee
485
Effect of long-term settlement on longitudinal mechanical performance of tunnel in soft soil

H.L. Zhao, X. Liu, Y. Yuan & Y. Chi
491
Theme 4: Safety issues, risk analysis harzard management and control

Research on stochastic seismic analysis of underground pipeline based on
physical earthquake model
X.Q. Ai & J. Li
499
Risk assessment for the safe grade of deep excavation

X.H. Bao & H.W. Huang
507
Multi-factors durability evaluation in subway concrete structure

C. Chen, L. Yang & C. Han
513
The use of artificial neural networks to predict ground movements caused by tunneling
I. Chissolucombe, A.P. Assis & M.M. Farias
519
Research and application of road tunnel structural optimization

W.Q. Ding & Y. Xu
525
VIII
Floor heave behavior and control of roadway intersection in deep mine

B.H. Guo & T.K. Lu
Squeezing potential of tunnels in clays and clayshales from normalized undrained
shear strength, unconfined compressive strength and seismic velocity
M. Gutierrez & C.C. Xia
531
537
Framework of performance-based fire protection design method for road tunnel

X. Han & G.Y. Ding
545
Prediction of surface settlements induced by shield tunneling: An ANFIS model

J. Hou, M.X. Zhang & M. Tu
551
Experimental studies of a geological measuring system for tunnel with ultrasonic transducer
D.H. Kim, U.Y. Kim, S.P. Lee, H.Y. Lee & J.S. Lee
555
Performance review of a pipe jacking project in Hong Kong

T.S.K. Lam
561
Geotechnical control of a major railway project involving tunnel works in Hong Kong
W. Lee, S.S. Chung, K.J. Roberts & P.L.R. Pang
567
Research on structural status of operating tunnel of metro in Shanghai and treatment ideas
J.P. Li, R.L. Wang & J.Y. Yan
573
Maximising the potential of strain gauges: A Singapore perspective

N.H. Osborne, C.C. Ng, D.C. Chen, G.H. Tan, J. Rudi & K.M. Latt
579
Discussion on design method for retaining structures of metro station deep excavations in Shanghai
R. Wang, G.B. Liu, D.P. Liu & Z.Z. Ma
587
Risk analysis for cutterhead failure of composite EPB shield based on fuzzy fault tree
Y.R. Yan, H.W. Huang & Q.F. Hu
595
Risk assessment on environmental impact in Xizang Road Tunnel

C.P. Yao, H.W. Huang & Q.F. Hu
601
Risk analysis and fuzzy comprehensive assessment on construction of shield

tunnel in Shanghai metro Line
H.B. Zhou, H. Yao & W.J. Gao
607
Theme 5: Physical and numerical modelling

Tunnel behaviour under seismic loads: Analysis by means of uncoupled and coupled approaches
D. Boldini & A. Amorosi
615
Investigating the influence of tunnel volume loss on piles using photoelastic techniques
W. Broere & J. Dijkstra
621
Assessment of tunnel stability in layered ground

P. Caporaletti, A. Burghignoli, G. Scarpelli & R.N. Taylor
627
Reinforcing effects of forepoling and facebolts in tunnelling

K. Date, R.J. Mair & K. Soga
635
Mechanical behavior of closely spaced tunnels laboratory model tests and FEM analyses
J.H. Du & H.W. Huang
643
Stability analysis of masonry of an old tunnel by numerical modelling and experimental design
J. Idris, T. Verdel & M. Alhieb
649
IX
Excavation with stepped-twin retaining wall: Model tests and numerical simulations
N. Iwata, H.M. Shahin, F. Zhang, T. Nakai, M. Niinomi & Y.D.S. Geraldni
655
Stability of an underwater trench in marine clay under ocean wave impact

T. Kasper & P.G. Jackson
663
A study on behavior of 2-Arch tunnel by a large model experiment

S.D. Lee, K.H. Jeong, J.W. Yang & J.H. Choi
669
Behavior of tunnel due to adjacent ground excavation under the influence of

pre-loading on braced wall
S.D. Lee & I. Kim
677
Two distinctive shear strain modes for pile-soil-tunnelling interaction in a granular mass
Y.J. Lee & C.S. Yoo
683
Stability analysis of large slurry shield-driven tunnel in soft clay

Y. Li, Z.X. Zhang, F. Emeriault & R. Kastner
689
Effects of soil stratification on the tunneling-induced ground movements

F.Y. Liang, G.S. Yao & J.P. Li
697
Centrifuge modelling to investigate soil-structure interaction mechanisms resulting from

tunnel construction beneath buried pipelines
A.M. Marshall & R.J. Mair
703
Ground movement and earth pressure due to circular tunneling: Model tests and
numerical simulations
H.M. Shahin, T. Nakai, F. Zhang, M. Kikumoto, Y. Tabata & E. Nakahara
709
Analysis of pre-reinforced zone in tunnel considering the time-dependent performance

K.I. Song, J. Kim & G.C. Cho
717
Vault temperature of vehicle fires in large cross-section road tunnel

K.S. Wang, X. Han & Z.X. Li
725
Effects of different bench length on the deformation of surrounding rock by FEM

X.M. Wang, H.W. Huang & X.Y. Xie
729
The effects of loaded bored piles on existing tunnels

J. Yao, R.N. Taylor & A.M. McNamara
735
3D FEM analysis on ground displacement induced by curved pipe-jacking construction

G.M. You
743
Theme 6: Calculation and design methods, and predictive tools

Calculation of the three dimensional seismic stressed state of Metro StationEscalatorOpen
Line Tunnels system, which is located in inclined stratified soft ground
R.B. Baimakhan, N.T. Danaev, A.R. Baimakhan, G.I. Salgaraeva, G.P. Rysbaeva,
Zh.K. Kulmaganbetova, S. Avdarsolkyzy, A.A. Makhanova & S. Dashdorj
751
A complex variable solution for tunneling-induced ground movements in clays

H.L. Bao, D.M. Zhang & H.W. Huang
757
Simulation of articulated shield behavior at sharp curve by kinematic shield model

J. Chen, A. Matsumoto & M. Sugimoto
761
Deformation and pore pressure model of the saturated silty clay around a subway tunnel
Z.D. Cui, Y.Q. Tang & X. Zhang
769
Analytical solution of longitudinal behaviour of tunnel lining

F.J.M. Hoefsloot
775
Design of tunnel supporting system using geostatistical methods

S. Jeon, C. Hong & K. You
781
Comparative study of software tools on the effects of surface loads on tunnels

D.K. Koungelis & C.E. Augarde
785
Geologic Model Transforming Method (GMTM) for numerical analysis modeling in

geotechnical engineering
X.X. Li, H.H. Zhu & Y.L. Lin
791
Review and interpretation of intersection stability in deep underground based

on numerical analysis
T.K. Lu, B.H. Guo, L.C. Cheng & J. Wang
799
Analysis of surface settlement due to the construction of a shield tunnel in soft

clay in Shanghai
Z.P. Lu & G.B. Liu
805
Urban tunnels in soil: Review of current design practice in Brazil

A. Negro
811
A study on loads from complex support system using simple 2D models

Z. Shi, W. Bao, J. Li, W. Guo & J. Zhu
817
Ground reaction due to tunnelling below groundwater table

Y.J. Shin, J.H. Shin & I.M. Lee
823
Basal stability of braced excavations in K0 -consolidated soft clay by upper bound method
X.Y. Song & M.S. Huang
829
Analytical two and three dimension models to assess stability and deformation magnitude of
underground excavations in soil
L.E. Sozio
837
Dynamic response of saturated silty clay around a tunnel under subway vibration loading
in Shanghai
Y.Q. Tang, Z.D. Cui & X. Zhang
843
Lateral responses of piles due to excavation-induced soil movements

C.R. Zhang, M.S. Huang & F.Y. Liang
Elastic-plastic analysis for surrounding rock of pressure tunnel with lining based on material
nonlinear softening
L.M. Zhang & Z.Q. Wang
849
857
Modification of key parameters of longitudinal equivalent model for shield tunnel

W. Zhu, X.Q. Kou, X.C. Zhong & Z.G. Huang
863
Author Index
869
XI
Preface
Under the Chairmanship of Professor K. Fujita, the first symposium purposely addressing geotechnical issues
related to underground construction in soft ground was held in 1994, prior to the 13th International Conference on
Soil Mechanics and Geotechnical Engineering held in New Delhi. Following the success of the first symposium,
Professor R. Mair succeeded the Chairmanship of TC28 and he initiated a series of three-day International
Symposia on Geotechnical Aspects of Underground Construction in Soft Ground including technical site visits
to underground construction projects. In total, four three-day International Symposia have been held very three
years since 1996. These include the ones held in London, UK (1996), in Tokyo, Japan (1999), in Toulouse, France
(2002) and in Amsterdam, the Netherlands (2005).
This volume includes a collection of four invited special lectures delivered by Dr A. Bezuijen
(The Netherlands), Mr Huang Rong (China), Professor M.D. Bolton (UK) and Professor I.M. Lee (Korea). The
titles of their lectures are Processes around a TBM, Overview of ShanghaiYangtze river tunnel project, Supporting excavations in clay from analysis to decision-making and Underground construction in decomposed
residual soils, respectively.
In addition, this volume contains 112 papers grouped under six themes including (i) Analysis and numerical
modelling of deep excavations; (ii) Construction method, ground treatment, and conditioning for tunnelling;
(iii) Case histories; (iv) Safety issues, risk analysis, hazard management and control; (v) Physical and numerical
modelling and (vi) Calculation and design methods, and predictive tools. Six general reports discussing and
commenting papers grouped under the six themes were contributed orally during the Symposium by Professor
Richard Finno, Professor Tadashi Hashimoto, Mr Alejo Sfriso, Dr C.T. Chin, Dr Richard Pang and Professor
Richard Kastner, respectively. The written versions of their six general reports are also included in this volume.
Y.S. Li
Chairman of the Symposium
C.W.W. Ng, H.W. Huang and G.B. Liu
Vice-Chairmen of the Symposium and Editors
XIII
Sponsors
Organized by:
Tongji University
Under the auspices of:
ISSMGE
Technical Committee 28 of the International

Society of Soil Mechanics and Geotechnical
Engineering (ISSMGE)
Supported by
China Civil Engineering Society
Chinese Society for Rock Mechanics

and Engineering
Geotechnical Division, the Hong Kong

Institution of Engineers
Hong Kong Geotechnical Society
Hong Kong University of Science

and Technology
XV
Science and Technology Commission of

Shanghai Municipality
Shanghai Changjiang Tunnel & Bridge
Development Co., Ltd.
Shanghai Society of Civil Engineering
XVI
Special lectures
Processes around a TBM

A. Bezuijen & A.M. Talmon
Deltares and Delft University of Technology, Delft, The Netherlands
ABSTRACT: Processes that occur around a TBM during tunnelling have been investigated while tunnelling in
saturated sand. The pore pressure in front of the TBM increases due to a lack of plastering during drilling. This
has consequences for the stability of the tunnel face, or the soil in front of the tunnel. A bentonite flow is likely
alongside the TBM from the tunnel face, and/or grout flow from the back. It seems that virtually no investigation
has been made of this part of the TBM, but it is important to understand the volume loss that occurs around a
tunnel. The lining is constructed behind the TBM and the tail void grout is applied. Pressures measured in the
tail void grout will be discussed, as well as the consequences for loading on the soil and the lining. Most of the
results described are based on field measurements performed at various tunnels constructed in the Netherlands.
INTRODUCTION
the current state of the art of these processes, and discusses how knowledge gained about these processes
may influence the design of a TBM tunnel in soft soil.
Dutch experience of using TBM tunnelling is relatively recent. The first TBM tunnel was constructed
in the Netherlands between 1997 and 1999 (the Second Heinenoord Tunnel). In the early 1990s, Dutch
engineers were uncertain whether the soft saturated
soil in the western parts of their country was suitable for TBM tunnelling. The decision was therefore
taken to include a measurement programme in the
first tunnelling projects. An overview of this programme and some results are presented by Bakker &
Bezuijen (2008). In the programme, results from the
measurements were predicted using existing calculation models. The measurement results were analysed
at a later date, and discrepancies with the predictions
were explained where possible.
An important part of the measurement and analysis programme was dictated by the processes that
occur around the TBM. This paper deals with some
of these processes. It does not cover all aspects of
TBM tunnelling as this would not fit within the limits
of this paper (see Bezuijen & van Lottum, 2006, for
more information). The paper focuses on certain areas
where ideas concerning the mechanisms involved have
changed over the last decade, and where a better
understanding is now apparent.
In order to structure this paper, we walk along
the TBM. We start with a process at the front of the
TBM: the creation and stability of the tunnel face under
the influence of excess pore pressures. The paper then
discusses what happens next to the TBM. The last
part of the paper deals with the tail void grout that
is injected at the end of the TBM. The paper describes
2
2.1
PORE PRESSURES IN FRONT OF A TBM

Flow in coarse and fine granular material
During TBM tunnelling, it is essential that the tunnel

face is stabilised by pressurised slurry (slurry shield)
or muck (EPB shield). The pressure must be adapted
to the ground pressure to stabilise the front. If pressure is too low, this will lead to an instable tunnel front
resulting in collapse of the tunnel face. If pressure is
too high, a blow-out will occur. Various calculation
methods have been proposed to calculate the stability of the tunnel face. Most of these methods do not
take the influence of pore water flow into account. It
is assumed that the bentonite slurry or muck at the
tunnel face creates a perfect seal that prevents water
flow from the face into the soil. Experience with tunnels built in areas where the subsoil contains gravel
has shown that the bentonite slurry can penetrate into
the subsoil over more than 7 m (Steiner, 1996). Steiner
advises that the sand and fines should be retained in
the slurry (instead of removing them in the separation
plant), and that sawdust should be used in the bentonite
(Steiner, 2007). Anagnostou & Kovari (1994) propose
a calculation method for such a situation. However, this
method only takes the viscous behaviour of the slurry
into account, and not the stiffening that occurs during standstill. The results of this calculation method
may therefore lead to the prescription of bentonite
Figure 1. Measured excess pore pressure in front of a slurry

shield and approximation.
Figure 2. Measured excess pore pressure in front of an

EPB shield () and approximation (Botlek Rail Tunnel, MQ1
South). Relatively impermeable subsoil.
with viscosity that is too high (Steiner, 2007). The

state of the art for such a situation involving coarse
granular material is still trial and error, but the trial
can be performed in the laboratory to avoid errors in
the field.
Usual tunnelling conditions in the Netherlands are
a saturated sandy soil in medium-fine sand. In such
soil conditions, the groundwater flow influences the
plastering. There will be virtually no plastering of
the tunnel face by the bentonite or the muck during
drilling, because the groundwater in front of the TBM
prevents water in the bentonite slurry or muck flowing into the soil. Plastering will only occur during
standstill of the TBM process.
Figure 1 shows measured pore pressure in front of
a slurry shield as a function of the distance from the
TBM front. Plastering occurs during standstill, resulting in a pressure of 120 kPa (the hydrostatic pressure).
Higher pore pressures were measured during drilling,
because the TBMs cutter head removes a cake before
it can form at the tunnel face.
Figure 2 shows the same phenomenon measured in
front of an EPB shield. Here, only the pressure during
drilling was recorded.
Bezuijen (2002) shows that the amount of excess
pore pressure measured in the soil in front of the
TBM (apart from pressure at the tunnel face) also
depends on soil permeability, the quality of the bentonite or muck, and the drilling speed. Where EPB
drilling takes place in sand with a low permeability
(k = 105 m/s), the pore pressure measured in sand in
front of the TBM is virtually equal to pressure in the
mixing chamber. The pressure is lower in sand with
higher permeability (k = 3 104 m/s), because some
plastering of the face occurs during drilling. Soil permeability also influences the foam properties. Muck in
the mixing chamber will be dryer in sand with a higher
permeability. Where the permeability of the sand is
lower, the water content in the muck is nearly entirely
determined by water in the soil and much less by the

foam properties (also see Bezuijen, 2002).
Figure 1 and Figure 2 also show a theoretical curve
(Bezuijen, 2002):
Where 0 is the piezometric head at the tunnel face,

the piezometric head at a distance x in front of
the tunnel face, and R the radius of the tunnel. This
relationship is valid for situations where the permeability of soil around the tunnel is constant. In the
Netherlands, the sandy layers used for tunnelling are
sometimes overlain with soft soil layers of peat and
clay with a low permeability. In such a situation, the
pressure distribution in the soil can be evaluated as
a semi-confined aquifer. This is described by Broere
(2001).
2.2 Influence on stability
Bezuijen et al (2001) and Broere (2001) have shown
that the groundwater flow in front of the TBM implies
that a larger face pressure is necessary to achieve a
stable front.According to Bezuijen et al (2001), the difference is approximately 20 kPa for a 10-m-diameter
tunnel constructed in sand, where the top is situated
15 m below the ground surface.
Knowledge of this groundwater flow appeared
essential during the Groene Hart Tunnel (GHT)
project, not to prevent collapse of the tunnel face but
to prevent a form of blow-out (Bezuijen et al, 2001).
This tunnel enters a deep polder where the piezometric
head in the sand layers underneath the soft soil layers is
higher than the surface level (see Figure 3). As a result,
the effective stresses beneath the soft soil layers are
extremely small. The calculated excess pore pressure
Table 1. Percentage of tapering of the TBM in 3 tunnel

projects in The Netherlands.
in the sand layer induced by the tunnelling process

could cause floating of the soft layers. The contractor made detailed numerical calculations (Aime et al,
2004). As a result of these calculations, a temporary
sand dam was constructed at the point where the tunnel
entered the polder. This dam delivered the necessary
weight to prevent lifting of the soft soil layers due
to excess pore pressure generated at the tunnel face
during drilling.
3.1
Tapering %
Second Heinenoord
Botlek
Sophia
0.95
0.77
0.79
the idea (Bezuijen, 2007) that the soil is not in contact with the TBM all over the TBM. Overcutting at
the tunnel face can lead to bentonite flow over the
TBM shield from the face towards the tail. Grout pressure during grout injection is usually higher at the tail
than the soil pressure. The soil is therefore pushed
away from the TBM, and grout will flow from the
tail over the shield. It is possible to describe flow on
the shield, if it is assumed that both the bentonite and
the grout are Bingham liquids, that the yield stress is
dominant in the flow behaviour, and that there is linear elastic soil behaviour. A more or less conceptual
model is developed, assuming a cylindrical symmetrical situation around the tunnel axis. Changes in the
soil radius for such a situation can be described as
(Verruijt, 1993):
Figure 3. Geotechnical profile GHT tunnel in polder. Tunnel is drilled from right to left in this picture.
Tunnel project
Where is the change in pressure, r the change

in radius, r the radius of the tunnel and the grout, and
G the shear modulus of the soil around the tunnel.
The flow around the TBM shield can be
described as:
FLOW AROUND THE TBM

Calculation model
Until recently, only limited attention has been given

to pressure distribution and flow around the TBM
shield. It was assumed that the soil was in contact
with the TBM shield across the shield. During drilling
of the Western Scheldt tunnel, however, it appeared
that the TBM deformed at large depths and high water
pressures (the tunnel is constructed up to 60 m below
the water line). This could not be explained by the
concept of a TBM shield in contact with the soil.
Furthermore, tunnelling technology has advanced to
a level where the ground loss due to tunnelling is
less than the volume difference caused by tapering of
the TBM. TBMs are usually tapered, with a slightly
larger diameter at the head compared with the tail.
This allows the TBM to manoeuvre and to drill with a
certain curvature. Table 1 shows the volume difference
due to tapering for different TBMs.
The volume losses measured during these projects
varied, but negative volume losses were sometimes
measured in all the projects (there was actually heave).
It is clear that the measured volume loss can be less
than the volume loss due to tapering. This leads to
Where P is the change in pressure due to the flow,

x a length increment along the TBM, s the gap width
between the tunnel and the soil, and the yield stress
of the grout around the TBM. is a coefficient indicating whether there is friction between the soil or
bentonite and the grout ( = 1) only, or also between
the TBM and the grout or bentonite ( = 2). Viscous
forces are neglected in this formula.This is permissible
due to the low flow velocities that can be expected.
With no grout or bentonite flow around the TBM,
tapering will lead to an effective stress reduction proceeding from the TBMs face to the tail according to
equation (2). The grout and bentonite flow will change
this pressure distribution. In order to calculate the pressure distribution under flow, the flow direction of both
the bentonite and the grout must be known. These
flow directions can vary during the tunnelling process (Bezuijen, 2007). On average, however, the TBM
Table 2. Input parameters used in calculation with bentonite

and overcutting.
Length TBM shield
Diameter
Diameter reduction
Overcutting
Asymmetric (1) or symmetric (2)
Grain stress
Grout pressure
Pore pressure
Pressure on tunnel face
Shear modulus (G)
Yield stress grout
Yield stress bentonite
5
10
0.2
0.015
2
150
400
200
250
90
1.6
0.01
The Figure shows that the gap width for a completely stiff soil mass would increase from 0.015 m at
the front to 0.025 m at the tail of the TBM. If there were
only grout pressures, the gap width would be 0.028 m
at the tail of the TBM, due to the grout pressure that
is larger than the total stress. However, the gap would
close at 3.4 m from the tail. If the influence of the bentonite is included, there is still a gap width of 0.01 m at
the tunnel face (5 m from the tail). The line through the
triangles presents the gap width due to the combined
effects of both the bentonite and the grout. The plot
above presents the pressures in the same way.
m
m
%
m
kPa
kPa
kPa
kPa
MPa
kPa
kPa
3.2 Consequences and status

The model shows that the volume loss is not determined by tapering of the TBM (as suggested for
example by Kasper & Meschke, 2006), but is influenced by the pressure distribution of the bentonite and
grout. With sufficient grout pressure, it is possible to
have a negative volume loss (the surface level rises
after the TBM passes). It also explains that bentonite is
sometimes found in the tail void, and grout is found in
the pressure chamber. The first situation occurs when
bentonite pressure is relatively high and grout pressure
is low (we will see that it is quite difficult to control
grout pressure, especially during ring building). The
second situation occurs when grout pressures in the
tail void are relatively high (which may occur during
drilling).
Contrary, however, to the model described for the
pore pressures in front of the TBM and the grout
pressure, to be described in the following sections, the
experimental evidence for this model is still limited. To
our knowledge, pressure distribution around the TBM
shield has never been measured. The shield was perforated during construction of the Western Scheldt tunnel
but no grout was found between the shield and the soil
(Thewes, 2007). The fact that no grout was found during this investigation may be caused by the fact that, in
reality, the TBM will not be placed as symmetrically in
the drilled hole as suggested in this simple model. The
TBM must be in contact with the soil at some point
to maintain mechanical equilibrium. There will be no
grout around the shield at that location.
Guglielmetti (2007) rightfully argues that more
research is needed in this field, because: The topic
(flow of bentonite and grout around the TBM) is
definitely one of the most important in the field of
mechanised tunnelling, being the management of the
void around the shield of a TBM as one of the major
sources of concern for both designers and contractors
involved in urban tunnelling projects.
There is some evidence from the results of extensometer measurements carried out at the Sophia Rail
Tunnel. The results of the extensometers (shown in
Figure 5) are presented in Figure 6 during passage of
Figure 4. Pressures and gap width along a TBM. Grout pressures and bentonite pressures. Parameters see Table 2. Plots
show pressures and gap width for the bentonite and grout
pressure separately and the combined result.
advances and therefore the bentonite and grout front

must also advance in the same direction to achieve a
stable situation. This means that grout and bentonite
only move with respect to the soil, and not with respect
to the TBM. Therefore = 1 for both the bentonite and
the grout. The result of an example calculation using
the parameters given in Table 2 is shown in Figure 4.
Figure 5. Sophia Rail Tunnel, soil stratification and location

of extensometers at the measurement location (picture Arne
Bezuijen).
Figure 7. First tube Sophia Rail Tunnel: drilling velocity

and measured grout pressures at the right side of the tunnel
as a function of time.
Figure 6. Extensometer results. The vertical line shows

when the tail of the TBM passes. Soil above the TBM is
already compressed before the tail passes.
Tunnel, as they have provided the most complete data

set until now.
The study of grout pressures was initiated by earlier
measurements performed at the Second Heinenoord
Tunnel and the Botlek Rail Tunnel. These measurements did not match the generally accepted assumption at that time at least in the Netherlands that the
vertical pressure gradient in liquid grout must be dictated by the density of the grout, and that the pressure
distribution after hardening must reflect the K0 (the
ratio between the horizontal and vertical soil pressure).
In reality, the vertical pressure gradient was lower and
the influence of K0 could not be detected.
the TBM. The results show that there is initially some

extension of the soil in front of the TBM due to the relatively low stresses at the tunnel face. However, the soil
above the tunnel (see the extensometer at 12.5 m) is
compressed several rings before the tail of the TBM
passes (the vertical line) indicating heave, and there
is therefore no settlement due to the tapering. When
the TBM has passed, the extensometer at 12.5 m follows the course of the grout pressures measured around
the lining. This will be discussed in more detail in the
next section, and shows that a change in grout pressure
indeed leads to a change in soil deformation.
We are currently working on the possibility of
measuring pressures around the shield.
4.2 Measurements
The Sophia Rail Tunnel was constructed in sandy subsoil overlain with soft soil layers (see Figure 5). The
water table is close to the surface. During construction of the Sophia Rail tunnel, two rings in the lining
were each equipped with 14 pressure sensors.The pressures measured with one of these instrumented rings
are shown in Figure 7.
These measurements are discussed in detail in
Bezuijen et al (2004): we will only describe the main
phenomena here. The upper plot in Figure 7 shows
the drilling velocity, when drilling occurs, and when
there was a standstill for ring building. It can be seen
that an increase in pressure is measured as soon as the
4 TAIL VOID GROUTING

4.1
Introduction tail void grouting
Coming at the end of the TBM, the tail void grouting process is important. The process determines the
loading on the soil and on the lining.
The pressure distribution caused by tail void grouting has been studied during construction of the Sophia
Rail Tunnel (Bezuijen et al, 2004) and the Groene
Hart Tunnel. Here, we will describe the fundamental
mechanisms using measurements from the Sophia Rail
pressure gauges (built into the lining elements) moved

from the grease into the grout. Pressure increases as
long as drilling continues, and decreases when drilling
stops during ring building.
4.3
Grout pressures
The mechanism that leads to these pressure variations is explained in Bezuijen & Talmon (2003). Grout
bleeding or consolidation of the grout leads to a
volume loss of grout. Experiments showed that this
volume loss is between 3% and 8%, depending on the
type of grout (Bezuijen & Zon, 2007). This consolidation leads to stress reduction in the relatively stiff sand
layer. This stress reduction is measured as a reduction
of grout pressure. The effective stresses will ultimately
be very small: the minimum stress that is necessary to
keep the hole in the ground open. Leca & Dormieux
(1990) calculate this for a tunnel opening in sand. They
calculate that a cylindrical cavity in the ground remains
open when effective stresses of only a few kPa are
applied.
The consequence is that grout pressures around the
lining will decrease to values that are only a few kPa
above the pore water pressure. It is therefore clear that
the original K0 can no longer be found in the grout pressures. The pressure decrease due to volume loss in the
grout has changed the original stress state, and unloading of the soil leads to much lower stresses. Since the
stresses in the sand around the tunnel decrease, the
sand reaction will be the reaction of a very stiff material. Only a small volume decrease in the grout will
lead to a large decrease in stresses. Calculation methods quite often still use the original in-situ stresses to
calculate loading on the lining. For a tunnel in sand,
this leads to a calculated loading that is much too high,
as shown by Hashimoto et al (2004).
For slow hardening or non-hardening grouts, the
strength increase in the grout is caused by grout bleeding or consolidation. It should be realised that this
strength increase is only present when the tunnel is
drilled through a permeable soil. When drilling takes
place through less permeable soils such as clay, this
consolidation will be much lower and the grout will be
in liquid form over a greater part of the tunnels length.
This has consequences for loading on the lining, as we
will discuss later.
Figure 8. First tube Sophia Rail Tunnel: pressure gradient

over the tunnel lining at one location, and pump activity for
one of the injection points (A1) as a function of time.
9.81 kPa/m. The tail void grout used for this tunnel
had a density of 2190 kg/m3 . If the vertical pressure
were to increase with depth in accordance with this
density, the pressure gradient should be 21.5 kPa/m.
Results showed that the measured vertical density is
always lower. This is because the grout is a Bingham
liquid, with a viscosity and a yield stress. The grout
has to flow downwards if more grout is injected in the
upper half of the tunnel. This downward flow needs
a driving force to overcome the yield stress, and the
pressure gradient will therefore be less than the gradient that is calculated from the density. Talmon et al
(2001) developed a numerical program to calculate the
pressure distribution in the tail void due to injection.
We only describe some of the consequences here. If the
viscosity is not taken into consideration, the maximum
pressure gradient (dP/dz) that can be expected is:
Where gr is the density of the grout, g the acceleration of gravity, the yield strength of the grout, and s
the width of the tail void gap between the soil and the
lining. If the yield stress in the grout is low, the vertical
pressure gradient is determined by the grout density
(21.5 kPa/m for the Sophia Rail Tunnel, slightly higher
than the maximum value measured in Figure 8). Consolidation or hardening of the grout leads to a higher
yield stress, and thus to a lower gradient.
A complicating factor is that the maximum shear
stress that can be developed is a vector. If the maximum shear stress is developed in one direction, there
will be no shear stress perpendicular to that direction. When drilling starts for a new ring and the grout
pumps are activated, the elastic soil reaction will lead
to an increase of the tail void and grout will therefore
flow backwards from the TBM. Ring shear stresses
barely develop in this situation, and the vertical gradients therefore increase during drilling. They decrease
again when drilling stops (Figure 8).
4.4 Pressure gradients

The vertical pressure gradient over the tunnel lining
is important when calculating the longitudinal loading
on the lining. The vertical pressure gradient that was
measured during construction of the first tunnel tube of
the Sophia Rail Tunnel is shown in Figure 8. The pressure gradient starts at nearly 20 kPa/m and decreases
to values under the pore water pressure gradient of
Further from the TBM, the vertical gradients

decrease and become equal to the gradient according to the buoyancy forces. This has to be the case,
because the total force on the lining far away from
the TBM must be zero. The vertical pressure gradient therefore compensates for the weight of the lining.
As a result, the gradient becomes lower than the gradient in the pore water. This is because the average
density of the lining is lower than the density of pore
water. One remarkable result is that the vertical pressure gradient at some distance from the TBM (at 12:00
in Figure 8, 5 rings behind the TBM) decreases during
drilling. The flow no longer has any influence at this
point, but drilling and grout injection lead to higher
gradients in the first part of the lining and therefore to
higher buoyancy forces. The first rings have the tendency to move upwards, which must be compensated
by the TBM and the rings further away. This partly
compensates for the weight of the rings further from
the TBM, so that the effective weight of these rings
and also the vertical gradient is less.
Figure 9. Position of pore pressure gauges and grout pressure gauges at ring 2117 of the GHT.
INFLUENCE ON PORE WATER PRESSURES
Section 2.1 describes how no plastering occurs at the

front when drilling takes place in fine to mediumfine saturated sand, because the bentonite filter cake is
destroyed by the cutting wheel before it is able to form.
As a result, water flows from the tunnel face into the
soil. Section 4.3 describes how consolidation of the
grout also leads to a water flow from the tunnel lining
into the soil, because water expelled from the grout
will flow into the surrounding soil. A grout cake will
form however, because the consolidated grout is no
longer disturbed. It is therefore reasonable to assume
that examination of the variation in pore pressure in
soil next to a tunnel under construction will show pore
pressures that are dominated by pressures existing at
the tunnel face. This theory was tested at the Groene
Hart Tunnel. Pore pressure transducers (PTTs) were
installed as close as 0.75 m from the tunnel lining. The
PTTs were placed in one plane, with the grout pressure
gauges on Ring 2117 of the tunnel (see Figure 9).
Figure 10 shows the measurement results. The grout
pressure gauges on Ring 2117 give no signal before
they are in the grout. The PPTs show a slight increase
during drilling due to the excess pore pressure generated at the tunnel face. As drilling stops, the pore
pressure reduces to the hydrostatic pressure. The various construction cycles can be seen. There is a sharp
increase in grout pressure when Ring 2117 leaves the
TBM, followed by a decrease due to consolidation. It
is remarkable however that this has virtually no influence on the measured pore pressures at less than a
metre from these gauges. This result is confirmed by
numerical calculations. The quantity of water expelled
Figure 10. Pore pressures and grout pressures measured at

GHT (also see text).
from the grout is far less than the water flow from the
tunnel face. The latter dominates the pore pressures.
The measurements show another remarkable feature. Grout pressure gauge 05 follows the water pressure after 3.20:00, but this is not the case for gauge 03.
This may indicate that there is no sealing grout layer
around gauge 05, so that it is possible to measure the
pore water pressure.
LOADING ON TUNNEL LINING
We have seen in Section 4.4 that vertical pressure gradients exist in the zone where the grout is not yet
consolidated or hardened which are higher than corresponds to the weight of the lining. Measurements at the
Sophia Rail Tunnel showed that the gradient decreases
more or less linear with the distance (see Figure 11).
As a result, that part of the lining is pressed upwards by
Figure 13. Boundary condition for beam calculation.

200
Analytical model
Bending moment [MNm]
150
Figure 11. Example of gradient in the grout pressure as a

function from the distance (0 on the X-axis represents the
point where the lining is more or less fixed. The TBM is at
9 m). Results from Sophia Rail Tunnel (Bezuijen et al, 2004).
Measurement Ring
2117
100
50
0
50
100
150
200
20
20
40
60
80 100 120
Disance behind TBM [m]
140
160
180
Figure 14. Bending moment ring 2117, measurement and

calculation. Groene Hart Tunnel (Hoefsloot, 2008).
However, Talmon (2007) has shown that such a

staged calculation is not necessary to find the same
results.
According to Talmon, the negative moment appears
at some distance from the TBM because the reaction
force to compensate the buoyancy in the fluid grout
zone is situated further from the TBM than the buoyancy force itself. The tunnel lining is pushed a bit
higher in the soil than in the equilibrium situation far
behind the TBM.
Hoefsloot and Talmon both model the tunnel lining
as a beam on an elastic foundation, except for lining
elements inside the TBM and lining elements in the
liquid grout zone, see Figure 13. The exact boundary
conditions and the transition between liquid and solid
grout are still the subject of debate.
Although example calculations have been presented
that show good correlation with measurements (see
Figure 14), there are still uncertainties with this type
of calculation that need further research:
Figure 12. Calculated shear force and moment in the lining, and displacement where the grout has not yet hardened.
Calculated moments are divided by 10.
the buoyancy forces. It is necessary to mobilise shear

forces from the TBM to achieve a stable tunnel lining.
This will lead to moments in the lining.
Bezuijen & Talmon (2005) have shown that the
moments in the liquid grout zone increase backwards
from the TBM (see Figure 12). A positive moment
means here that the force on the lower part of the tube
is higher than on the upper part. At the TBM, this
moment is created by the TBM itself. This is because
face pressure is higher at the bottom due to larger soil
stresses.
At the Groene Hart Tunnel the bending moment in
the lining was measured for a large distance behind
the TBM using strain gauges installed in the lining
segments. There is an increase in the moment for a few
rings, in accordance with the calculations previously
mentioned. There is subsequently a decrease, with the
moments becoming negative at a greater distance from
the tunnel. Bogaards & Bakker (1999) and Hoefsloot
(2008) argue that the remaining bending moment is a
result of the staged construction of the tunnel. They
developed a calculation model to take into account the
different stages in construction.
An important input parameter is the moment and

shear force that is transferred from the TBM to the
lining. While the moment can be derived from the
jack forces, the shear force is not determined.
With generally-accepted parameters for the lining
stiffness and the soils elastic parameters, the calculated movement of the lining is much smaller than
the measured movement.
The grout pressures are only measured when the
grout is more or less in the liquid phase. This results
10
Table 3. Specification of grout mixtures used in fracture

tests (WCR = water-cement ratio). Coclay D90 Ca activated
bentonite is used.
Mixture
1
2
WCR
Bentonite
%
k
(m/s)
1
10
7
7
5.108
6.100
in loading on the tunnel lining as shown in Figure 11. However, loading on the lining in situations
where the grout has hardened is less known. This is
because the instruments used were not suitable to
measure pressures when grout has hardened.
Conclusions that can be drawn from this type of
calculations are:
Figure 15. Density measured with a CT-scan. Raw data

(inset) and density. Correction for beam hardening effect and
calculated value of the density of the grout along the line
shown in inset. Mixture 1 in Table 3.
The length of the liquid grout zone and the density

of the grout are extremely important parameters
when calculating bending moments in the lining. If
this length is too long, loading will be too high and
tunnelling will not be possible (also see Bezuijen &
Talmon, 2005).
The shear force that is exerted on the lining by
the TBM is an important parameter. It is therefore
worthwhile to measure this shear force.
COMPENSATION GROUTING
Grout consolidation also appeared to be important when describing compensation grouting. Experiments (Gafar et al, 2008) showed that the fracturing
behaviour in compensation grouting depends on the
specification of the grout. If more cement is added,
the permeability of the grout is higher and there will
be more consolidation and leak-off during grout injection. Gafar et al describe how this influences the
fracturing behaviour. Recent tests carried out as part of
the research project on compensation grouting present
proof of the suggested grout consolidation mechanism.
At Delft University, the density of grout bodies made in
two compensation grouting experiments was analysed
in a CT-scan. Such a CT-scan can be used to determine
the density of the material tested. The grout mixtures
used in the experiments are shown in Table 3.
The results of the CT-scans are shown in Figure 15
and Figure 16. The results of the first grout mixture
clearly show an increase in density at the boundary of
the grout body. Grout at the boundary of the sample
is consolidated. The grout body made with the second
mixture has a more constant density across the fracture (the middle section). In the second experiment,
the CT-scan was performed while the grout body was
still in the sand. The more homogeneous density of
Figure 16. Grout density in a fracture measured with a

CT-scan. Mixture 2 in Table 3.
the grout body in the second test is understandable

if the permeabilities of the grout are considered. The
lower permeability of the second grout sample results
in much less grout consolidation within the limited
injection time. The grout density in the fracture therefore does not increase at the boundary of the grout as
is the case for mixture 1.
The permeabilities were determined using the procedure suggested by McKinley and Bolton (1999), a
form of oedometer test with drainage on one side. This
procedure can also be used to test the consolidation
properties of tail void grout. However, the thickness
of the grout layer in the test should be identical to that
in the field. This is to avoid scaling effects that occur
11
A conceptual model for the flow of bentonite and

grout has been developed. Although this model
must still be verified using the results of measurements, it shows some promising results.
Considerable information has been obtained about
the grouting process and the resultant lining
loading.
because hardening of the grout is independent of the

sample size (Bezuijen & Zon, 2007).
DISCUSSION
The research described above has increased understanding of the processes that occur around a TBM
during tunnelling. This has already had consequences
for practical aspects of tunnelling. Examples are the
excess pore pressures in front of the TBM: extra sand
was added locally above the planned tunnel trajectory of the Groene Hart Tunnel to prevent a blow-out,
and the grout was changed in a tunnel project in
London where it appeared that the liquid zone of traditional grout for a tunnel drilled in clay with no
possibility of consolidation was too long to achieve
the desired drilling speed. However, the authors believe
that the results can make an even greater contribution
to improving shield tunnelling. Knowledge about the
influence of excess pore pressures on face stability can
improve definition of the pressure window at the tunnel face, so preventing a blow-out due to excessively
high pressures and instability caused by pressures that
are too low. In combination with research on EPB tunnelling in clay (Merrit & Mair, 2006), foam research
for EPB tunnelling in sand can lead to better control
of the EPB process. It has already been discussed how
flow around theTBM is important forTBM design, and
that more experimental evidence is needed. Research
into grouting can lead to smaller settlement troughs
and optimisation of loading on the lining. This last
aspect may lead to cheaper lining construction.
The results must be discussed with tunnel builders
and contractors if improvements to the shield tunnelling process are to be achieved. Discussion about
certain aspects has already started, but we hope that
this paper will stimulate the involvement of more
parties.
Although not unusual, it is interesting to see that

this research also raises new questions: what is the
exact position of the TBM during the tunnelling process, what is the interaction between the TBM and the
lining, are the predicted pressures around the TBM
correct, and what are the consequences for our design
methods? Even in a relatively simple beam calculation
for calculating loading on the lining in a longitudinal
direction it appears that uncertainties in the boundary
conditions determine the outcome of the calculation.
As long as these uncertainties remain, more sophisticated numerical calculations will present the same
uncertainties.
ACKNOWLEDGEMENTS
The research described in this paper was sponsored by
COB, the Dutch Centre for Underground Construction, and Delft Cluster. We would like to thank these
organisations for giving us the opportunity to perform
this research. We also wish to thank the project organisations of the different tunnels for giving permission
to use tunnelling data in our research. And last but not
least, we would like to thank our fellow members in
the COB committees for their stimulating discussions
on the various subjects.
REFERENCES
Aime, R., Aristaghes, P., Autuori, P. & Minec, S. 2004. 15 m
Diameter Tunneling under Netherlands Polders. Proc.
Underground Space for Sustainable Urban Development
(ITA Singapore), Elsevier.
Anagnostou, G. & Kovri, K. 1994. The face stability of
Slurry-shield-driven Tunnels. Tunelling and Underground
Space Technology 9(2): 165174.
Bakker, K.J. & Bezuijen, A. 2008. 10 years of bored tunnels in the Netherlands. Proceeding 6th Int. Symposium
on Underground Construction in soft Ground, Shanghai.
Bezuijen, A. 2002. The influence of soil permeability on the
properties of a foam mixture in a TBM. 3rd. Int. Symp.
on Geotech. Aspects of Underground Construction in Soft
Ground, IS-Toulouse.
Bezuijen, A. 2007. Bentonite and grout flow around a TBM.
Proc. ITA 2007, Prague.
Bezuijen, A., Pruiksma, J.P. & Meerten, H.H. van. 2001. Pore
pressures in front of tunnel, measurements, calculations
and consequences for stability of tunnel face. Proc. Int.
Symp. on Modern Tunneling Science and Techn. Kyoto.
CONCLUSIONS
To understand the processes that are important when

tunnelling with a TBM, the flow processes around a
TBM must be considered: groundwater flow at the tunnel face, bentonite and grout flow around theTBM, and
grout flow and grout consolidation around the tunnel
lining. The research described in this paper has brought
about progress with regard to these flow processes
during tunnelling in soft ground:
The groundwater flow at the tunnel face is
described.
The muck in the mixing chamber is described as a
function of drilling speed and permeability.
12
Bezuijen, A. & Talmon, A.M. 2003. Grout the foundation of

a bored tunnel. Proc ICOF 2003 Dundee.
Bezuijen,A., Talmon,A.M., Kaalberg, F.J. & Plugge, R. 2004.
Field measurements of grout pressures during tunneling
of the Sophia Rail tunnel. Soils and Foundations 44(1):
4150.
Bezuijen, A. & van Lottum, H. (eds). 2006. Tunnelling A
Decade of Progress. GeoDelft 1995-2005, Taylor and
Francis/Balkema, Leiden, ISBN 0 415 39113 4.
Bezuijen, A. & Zon, W. van der. 2007. Volume changes in
grout used to fill up the tail void. Proc. ITA 2007, Prague.
Bogaards, P.J. & Bakker, K.J. 1999, Longitudinal bending
moments in the tube of a bored tunnel. Numerical Models
in Geomechanics Proc. NUMOG VII: 317321.
Broere, W. 2001. Tunnel Face Stability & New CPT Applications. Ph.D. Thesis, Delft University of Technology, Delft
University Press.
Gafar, K., Soga, K., Bezuijen, A., Sanders, M.P.M. & van
Tol, A.F. 2008. Fracturing of sand in compensation grouting. Proceeding 6th Int. Symposium on Underground
Construction in soft Ground, Shanghai.
Guglielmetti, V. 2007. Tunnels and Tunnelling International,
October, P32.
Hashimoto, T., Brinkman, J., Konda, T., Kano, Y. &
Feddema, A. 2004. Simultaneous backfill grouting, pressure development in construction phase and in the long
term. Proc. ITA Singapore.
Hoefsloot, F.J.M. 2008. Analytical solution longitudinal
behaviour Tunnel lining. Proceeding 6th Int. Symposium
Kasper, T. & Meschke, G. 2006. On the influence of face

pressure, grouting pressure andTBM design in soft ground
tunnelling. Tunn. and Undergr. Space Techn. 21: 160171.
Leca, E. & Dormieux, L. 1990. Upper and lower bound solutions for the face stability of shallow circular tunnels in
frictional material. Gotechnique 43: 519 (in French).
Merritt, A.S. & Mair, R.J. 2006. Mechanics of tunnelling
machine screw conveyors: model tests. Geotechnique
56(9): 605615.
McKinley, J.D. & Bolton, M.D. 1999. A geotechnical description of fresh cement grout Filtration and consolidation behaviour. Magazine of Concrete Research 51(5):
295307.
Steiner, W. 1996. Slurry penetration into coarse grained
soils and settlements from a large slurry shield tunnel.
Proc. Geotech. Aspects of Underground Construction in
Soft Ground, London, Mair and Taylor (eds). Balkema,
Rotterdam, ISBN 9054108568: 329333.
Steiner, W. 2007. Private communication.
Talmon, A.M. 2007. Notes on analytical beam model. Delft
Hydraulics report Z3934/Z4145.
Talmon, A.M., Aanen, L. Bezuijen, A & Zon, W.H. van der.
2001. Grout pressures around a tunnel lining Proc. Int.
Symp. on Modern Tunneling Science and Techn. Kyoto.
Thewes, M. 2007. Private communication.
Verruijt, A. 1993. Soil Dynamics. Delft University of Technology, b28.
13
Supporting excavations in clay from analysis to decision-making

M.D. Bolton & S.Y. Lam
University of Cambridge, UK
A.S. Osman
Durham University, UK
ABSTRACT: Finite Element Analysis (FEA) is used to calibrate a decision-making tool based on an extension
of the Mobilized Strength Design (MSD) method which permits the designer an extremely simple method of
predicting ground displacements during construction. This newly extended MSD approach accommodates a
number of issues which are important in underground construction between in-situ walls, including: alternative
base heave mechanisms suitable either for wide excavations in relatively shallow soft clay strata, or narrow
excavations in relatively deep soft strata; the influence of support system stiffness in relation to the sequence of
propping of the wall; and the capability of dealing with stratified ground. These developments should make it
possible for a design engineer to take informed decisions on the relationship between prop spacing and ground
movements, or the influence of wall stiffness, or on the need for and influence of a jet-grouted base slab, for
example, without having to conduct project-specific FEA.
INTRODUCTION
(Osman et al. 2006) and also the sequential construction of braced excavations which induce wall
displacements and ground deformations (Osman and
Bolton, 2006).
Consider the imposition of certain actions on a soil
body, due to construction activities such as stress relief
accompanying excavation, or to loads applied in service. The MSD method permits the engineer to use
simple hand calculations to estimate the consequential ground displacements accounting for non-linear
soil behavior obtained from a single well-chosen test
of the undisturbed soil.
The MSD approach firstly requires the engineer to
represent the working states of the geotechnical system
by a generic mechanism which conveys the kinematics
(i.e. the pattern of displacements) of the soil due to the
proposed actions. Analysis of the deformation mechanism leads to a compatibility relationship between the
average strain mobilized in the soil and the boundary
displacements.
The average shear strength mobilized in the soil due
to the imposed actions is then calculated, either from an
independent equilibrium analysis using a permissible
stress field (equivalent to a lower bound plastic analysis), or from an equation balancing work and energy for
the chosen mechanism (equivalent to an upper bound
plastic analysis).
The location of one or more representative soil elements is then selected, basing this judgment on the
soil profile in relation to the location and shape of
The Mobilizable Strength Design (MSD) method has

developed following various advances in the use of
plastic deformation mechanisms to predict ground displacements: (Milligan and Bransby, 1975; Bolton and
Powrie, 1988; Bolton et al. 1989, 1990a, 1990b). MSD
is a general, unified design methodology, which aims
to satisfy both safety and serviceability requirements
in a single calculation procedure, contrasting with conventional design methodology which treats stability
problems and serviceability problems separately. In
the MSD method, actual stress-strain data is used to
select a design strength that limits ground deformations, and this is used in plastic soil analyses that satisfy
equilibrium conditions without the use of empirical
safety factors.
Simple plastic mechanisms are used to represent the
working state of the geotechnical system. The mechanisms represent both the equilibrium and deformation
of the various soil bodies, especially at their junction
with the superstructure. Then, raw stress-strain data
from soil tests on undisturbed samples, taken from representative locations, are used directly to link stresses
and strains under working conditions. Constitutive
laws and soil parameters are unnecessary.
The MSD approach has been successfully implemented for shallow foundations (Osman and Bolton,
2005), cantilever retaining walls (Osman and Bolton,
2005), tunneling-induced ground displacements
15
considering an admissible plastic mechanism for base

heave. In this case, the mobilized shear strength was
deduced from the kinematically admissible mechanism itself, using virtual work principles. The energy
dissipated by shearing was said to balance the virtual
loss of potential energy due to the simultaneous formation of a subsidence trough on the retained soil
surface and a matching volume of heave inside the
excavation. The mobilized strength ratio could then be
calculated, and the mobilized shear strain read off from
the stress-strain curve of a representative element, as
before. The deformation is estimated using the relationship between the boundary displacements and the
average mobilized shear strain, in accordance with the
original mechanism.
The MSD solutions of Osman and Bolton (2006)
compared quite well with some numerical simulations
using the realistic non-linear MIT-E3 model, and various case studies that provided field data. However,
these initial solutions are capable of improvement in
three ways that will contribute to their applicability in
engineering practice.
the selected mechanism. The centroid of the mechanism can serve as a default location if a single
location is to be employed. Stress-strain relationships
are then obtained from appropriate laboratory tests
on undisturbed soil samples taken from the selected
locations and carried out with precise strain measurements. Equivalent in-situ tests such as self-boring
pressuremeter tests can alternatively be carried out.
The mode of deformation in the soil tests should correspond as closely as possible to the mode of shearing
in the MSD mechanism. Otherwise, anisotropy should
somehow be allowed for.
Finally, the mobilized shear strength required for
equilibrium under working loads is set against the representative shear stress-strain curve in order to obtain
the mobilized soil strain, and thereby the boundary
displacements of the simplified MSD mechanism.
2
MSD FOR DEEP EXCAVATION PROBLEM
Osman and Bolton (2006) showed for an in-situ wall

supporting a deep excavation that the total deformation could be approximated as the sum of the
cantilever movement prior to propping, and the subsequent bulging movement that accretes incrementally
with every sequence of propping and excavation.
A method for estimating the cantilever movement
had been suggested earlier in Osman and Bolton
(2004). It begins by considering the lateral earth pressure distribution for a smooth, rigid, cantilever wall
rotating about a point some way above its toe, in
undrained conditions. A simple mobilized strength
ratio is introduced to characterize the average degree
of mobilization of undrained shear strength throughout the soil. By using horizontal force and moment
equilibrium equations, the two unknowns the position of the pivot point and the mobilized strength
ratio are obtained. Then, a mobilized strain value
is read off from the shear stress-strain curve of a soil
element appropriate to the representative depth of the
mechanism at the mid-depth of the wall. Simple kinematics for a cantilever wall rotating about its base
suggests that the shear strain mobilized in the adjacent
soil is double the angle of wall rotation. Accordingly,
for the initial cantilever phase, the wall rotation is estimated as one half of the shear strain required to induce
the degree of mobilization of shear strength necessary
to hold the wall in equilibrium. Osman and Bolton
(2004) used FEA to show that correction factors up to
about 2.0 could be applied to the MSD estimates of
the wall crest displacement, depending on a variety of
non-dimensional groups of parameters ignored in the
simple MSD theory, such as wall flexibility and initial
earth pressure coefficient prior to excavation.
A typical increment of bulging, on the other hand,
was calculated in Osman and Bolton (2006) by
1 The original mechanism assumed a relatively wide

excavation, whereas cut-and-cover tunnel and subway constructions are likely to be much deeper than
their width. The MSD mechanism therefore needs
to be adapted for the case in which the plastic deformation fields for the side walls interfere with each
other beneath the excavation.
2 The structural strain energy of the support system can be incorporated. This could be significant
when the soil is weak, and when measures are
taken to limit base heave in the excavation, such
as by base grouting between the supporting walls.
In this case, the reduction of lateral earth pressure due to ground deformation may be relatively
small, and it is principally the stiffness of the
structural system itself that limits external ground
displacements.
3 Progressively incorporating elastic strain energy
requires the calculation procedure to be fully incremental, whereas Osman and Bolton (2006) had
been able to use total energy flows to calculate
the results of each stage of excavation separately.
A fully incremental solution, admitting ground layering, will permit the accumulation of different
mobilized shear strengths, and shear strains, at
different depths in the ground, thereby improving
accuracy.
It is the aim of this paper to introduce an enhanced
MSD solution that includes these three features. This
is then compared with existing FEA of braced excavations which featured a range of geometries and
stiffnesses. It will be suggested that MSD provides
the ideal means of harvesting FEA simulations for use
in design and decision-making.
16
PLASTIC FAILURE MECHANISMS
Limit equilibrium methods are routinely used in stability calculations for soft clay which is idealised, unrealistically, as rigid-plastic. Slip surfaces are selected
as the assumed focus of all plastic deformations. Failure mechanisms should be kinematically admissible,
meaning that unwanted gaps and overlaps should not
be produced. Furthermore, in the case of undrained
shearing of clays, a constant-volume condition should
be respected at every point. A consequence is that
undrained plane-strain failure mechanisms must comprise only slip planes and slip circles. The soil on such
failure surfaces is taken to mobilize its undrained shear
strength divided by a safety factor, to maintain the
mechanism in limiting equilibrium under the action
of gravity, and any other applied loads. Calculated in
this way, the safety factor literally offers an estimate of
the factor by which the strength of the soil would have
to drop before the soil construction would collapse.
Such estimates might err either on the high side or the
low side, depending on the particular assumptions that
were made.
In the case of base heave in braced excavations, plastic solutions were derived from slip-line fields based
on the method of characteristics. Such solutions comprise both slip surfaces, as before, and plastic fans
which distribute plastic strains over a finite zone in
the shape of a sector of a circle. Notwithstanding
these zones of finite strain, the additional presence
of slip surfaces still restricts the application of these
solutions to the prediction of failure. Furthermore,
no such solution can be regarded automatically as
an accurate predictor of failure, notwithstanding their
apparent sophistication. All that can be said is that
they will lead to an unsafe estimate of stability. Their
use in practice can only be justified following backanalysis of actual failures, whether in the field or the
laboratory.
Two typical failure mechanisms as suggested by
Terzaghi (1943) and Bjerrum and Eide (1956) are
shown in Figure 1. They have each been widely used
for the design of multi-propped excavations. Terzaghi (1943) suggested a mechanism consisting of a soil
column outside the excavation which creates a bearing
capacity failure. The failure is resisted by the weight
of a corresponding soil column inside the excavation
and also by adhesion acting along the vertical edges
of the mechanism. Bjerrum and Eide (1956) assumed
that the base of the excavation could be treated as a
negatively loaded perfectly smooth footing. The bearing capacity factors proposed by Skempton (1951) are
used directly in the stability calculations and are taken
as stability numbers, N = H/cu . Eide et al. (1972)
modified this approach to account for the increase in
basal stability owing to mobilized shear strength along
the embedded length of the rigid wall.
Figure 1. Conventional basal stability mechanism and notation (after Ukritchon et al. 2003).
ORourke (1993) further modified the basal stability calculations of Bjerrum and Eide (1956) to include
flexure of the wall below the excavation level. It was
assumed that the embedded depth of the wall does
not change the geometry of the basal failure mechanism. However, an increase in stability was anticipated
due to the elastic strain energy stored in flexure. This
gave stability numbers that were functions of the yield
moment and assumed boundary conditions at the base
of the wall.
Ukritchon et al. (2003) used numerical limit analysis to calculate the stability of braced excavations.
Upper and lower bound formulations are presented
based on Sloan and Kleeman (1995) and Sloan (1988),
respectively.The technique calculates upper bound and
lower bound estimates of collapse loads numerically,
by linear programming, while spatial discretization
and interpolation of the field variables are calculated using the finite element method. No failure
17
For short walls embedded in deep soft clay, the maximum wall displacement occurs at the tip of the wall so
the wavelength was taken as twice the projecting wall
length ( = 2). Intermediate cases might be described
as restrained-end walls (1 < < 2).
However, these definitions applied only to very
wide excavations. When a narrow excavation is considered, the wavelength will be limited by the width
of the excavation. In addition, in the case of the partially restrained wall, the depth of a relatively stiff soil
stratum may also limit the depth of the deformation
pattern.
Figure 2. Incremental displacements in braced excavation

(after ORourke, 1993).
mechanism need be assumed and failure both of the

soil and the wall are taken care of. However, both soil
and wall are again assumed to be rigid perfectly plastic so the failure mechanism includes a plastic hinge
at the lowest level of support.
All these collapse limit analyses provide useful
guidance on the possible geometry of plastic deformation mechanisms for service conditions. But the key
requirement for MSD mechanisms is that displacement discontinuities (slip surfaces) must be avoided
entirely. In that way, small but finite ground displacements are associated at every internal point with small
but finite strains.
GEO-STRUCTURAL MECHANISMS
An incremental plastic deformation mechanism conforming to Equation 1 was proposed by Osman and
Bolton (2006) for an infinitely wide multi-propped
excavation in clay. In this mechanism, the wall is
assumed to be fixed incrementally in position and
direction at the lowest prop, implying that the wall has
sufficient strength to avoid the formation of a plastic hinge. The wall and soil are deforming compatibly
and the soil deformation also follows the cosine function of Equation 1. The dimensions of this mechanism
depend on the wavelength l.
Figure 3(a) shows the complete displacement field
for the mechanism proposed by Osman and Bolton
(2006). The solution includes four zones of distributed
shear which consist of a column of soil adjoining the
excavation above the level of the lowest prop, a circular
fan zone centred at the lowest prop, another circular fan
zone with its apex at the junction of the wall and the
excavation surface and a 45 degree isosceles wedge
below the excavation surface. It is required that the
soil shears compatibly and continuously with no relative sliding at the boundaries of each zone. The dotted
lines with arrows show the direction of the flow. Along
each of these lines the displacement is constant and
is given by the cosine function of Equation 1. It is
assumed that the zone outside the deformation zones
is rigid. This mechanism is simple and neat, but it
only applies to very wide excavations. In the case of a
narrow excavation, the width of the triangular wedge
could be bigger than the actual width of the excavation.
In view of this, a new mechanism for narrow excavations is proposed in Figure 3(b). The mechanism in
the passive zone (zone EFHI) is replaced. The new
mechanism meets the condition for undrained shearing, which means that the volumetric strain remains
zero throughout the zone.
The following solution approach is an extension
of Osman and Bolton (2006). In their original solution, soils are assumed to be homogenous. The average
shear strain increment in each zone is calculated by
taking the derivative of the prescribed displacement
4 WALL DEFORMATIONS
Consider now the deformations of a multi-propped
wall supporting a deep excavation in soft, undrained
clay. At each stage of excavation the incremental displacement profile (Figure 2) of the ground and the wall
below the lowest prop can be assumed to be a cosine
function (ORourke, 1993) as follows:
Here w is the incremental wall displacement at any

distance y below the lowest support, wmax is its
maximum value, and l is the wavelength of the deformation, regarded as proportional to the length s of the
wall below the lowest level of current support:
ORourke (1993) defined the wavelength of the deformation as the distance from the lowest support level to
the fixed base of the wall. Osman and Bolton (2006)
suggested a definition for the wavelength of the deformation based on wall end fixity. For walls embedded
into a stiff layer beneath the soft clay, such that the
wall tip is fully fixed in position and direction, the
wavelength was set equal to the wall length ( = 1).
18
equation. Then, the undrained shear strength (cu,mob )

mobilized at any location for any excavation heightwas expressed using a single mobilization ratio
( = cu,mob /cu ) to factor the strength profile. With the
use of the virtual work principle, the plastic work
done by shearing of the soil was equated to the virtual
change of gravitational potential energy of the soil.
A factor can then be found so that a corresponding mobilized shear strain can be read off from the
chosen stress-strain curve. The incremental displacement can then be calculated by the correlation between
the average shear strain increment and the incremental
wall displacement.
This approach offered a straightforward way to estimate the bulging displacement of the retaining wall.
However, the approach requires refinement in order to
include some additional features that may be significant in deep excavations. Firstly, the approach did not
consider the elastic strain energy stored in the support
system. Secondly, it is common to find a non-uniform
soil stratum with undrained shear strength varying
irregularly with depth. Furthermore, the geometry of
the deformation mechanism changes as the construction proceeds, so the representation of mobilization
of shear strength through the whole depth, using a
single mobilization ratio, is only a rough approximation. In reality there will be differences in mobilization
of shear strength at different depths for calculating
incremental soil displacement. Lastly, the original
mechanism of Osman and Bolton (2006) shown in
Figure 3(a) only applied to wide excavations; narrow excavations called for the development of the
alternative mechanism of Figure 3(b).
In view of these issues, a new fully incremental calculation method has been introduced, allowing for the
storage of elastic strain energy in the wall and the support system, and respecting the possible constriction
of the plastic deformations due to the narrowness of
an excavation.
dwmax
D
F
dwmax
Excavation
depth
dwmax
dwmax
H
E
ave 2
wmax
l
t
Hard stratum
(a) Incremental displacementfield for wide excavation

l
ave 2 wmax
l
dwmax
ave 2.2 wmax

l
Excavation
depth
dwmax
dwmax
t
Hard stratum
I
5
4
3
5.1
Deformation pattern in different zones
2
1
From Figure 3, the soil is assumed to flow parallel

to the wall at the retained side above the level of the
lowest support (zone ABDC) and the incremental displacement at any distance x from the wall is given by
the cosine function of Equation 1, replacing y by x.
By taking the origin as the top of the wall, the
deformation pattern of retained soil ABDC is given
in rectangular coordinates as follows:
0
1
2
3
4
5
0
9 10
(b) Incremental displacement field for narrow excavation
Figure 3. Incremental displacement fields.
19
In fan zone, CDE, by taking the apex of the fan zone

as the origin
A
Layer 1
Layer 2
Layer 3
Layer 4
B
s1
DSS
D
s1
For fan zone EFH in very wide excavations as indicated in Figure 3(a), by taking the junction of the wall
and the current excavation level as the origin:
I
PSP
s1
PSA
s1
DSS
Layer (n-1)
Layer n
Figure 4. Mobilizable shear strength profile of an excavation stage in an layered soil.
For the triangular zone FHI in very wide excavations, again taking the junction of the excavation and
the wall as the origin:
be obtained. On the active side of the excavation, the

spatial scale is fixed by the wavelength of deformation l, and all strain components are proportional to
dwmax /l. The average engineering shear strain increment mob mobilized in the deformed soil can be
calculated from the spatial average of the shear strain
increments in the whole volume of the deformation
zone. For a wide excavation i.e. Figure 3(a), the average
shear strain is equal to 2dwmax /l. For a narrow excavation, the average shear strain ( ave ) of active zone
ABCD and fan zone CDE is 2dwmax /l and 2.23dwmax /l,
respectively, while ave in the passive zone EFHI
depends both on the wavelength l of the deformation
and the width B of the excavation. The relationship
between the normalized average shear strain in EFHI
and the excavation geometry is shown in Figure 5. The
correlations are as follows:
For narrow excavations as shown in Figure 3(b), a

rectangular zone EFHI of 2D shearing is now proposed. The origin is taken as the mid-point of FE,
mid-wavelength in the excavation, at the wall.
In order to get more accurate solutions, it is supposed that the soil stratum is divided into n layers
of uniform thickness t (Figure 4). The average shear
strain d(m,n) is calculated for n layers in m excavation stages. The incremental engineering shear strain
in each layer is calculated as follows:
Apart from the first excavation stage, all subsequent

deformation mechanisms must partially overlay the
previous ones (Figure 6). Due to the non-linearity of
soil it is important to calculate the accumulated mobilized shear strain in each particular layer of soil in
order to correctly deduce the mobilized shear strength
of that layer. This is done by an area average method
described as follows:
In order to get a better idea of the deformation

mechanism, the relationship between the maximum
incremental wall displacement and the average shear
strain mobilized in each zone of deformation should
20
Figure 7. Typical stress strain relationship of soft clay.

Figure 5. Correlation between normalized average shear
strain and excavation geometry for a narrow excavation.
5.2 Shear strength mobilized in mechanism

In soft clay, the undrained shear strength generally
varies with depth, and with orientation of shear direction. The strength matrix cmob (m,n) mobilized for
excavation stage m for layer n can be expressed using
a matrix (m,n) on the appropriate undrained shear
strength profile. Regarding orientation, anisotropy of
soft soil can be a significant factor for excavation stability. For example, Clough & Hansen (1981) show an
empirical factor based on the observation that triaxial
extension tests give roughly one half the undrained
shear strength of triaxial compression, with simple
shear roughly half way between. Figure 4 shows the
orientation of the major principal stress direction
within the various zones of shearing in the assumed
plastic mechanism for wide excavations, and indicates
with a code the soil test configuration that would correctly represent the undrained shear strength of at the
specific orientation. For locations marked DSS the
assume directions of shearing are either vertical or
horizontal, so the ideal test on a vertical core is a
direct simple shear test. In areas marked PSA and
PSP, shearing takes place at 45 degrees to the horizontal and these zones are best represented by plain
strain active and passive tests, respectively. Since the
undrained shear strength of the direct simple shear test
is roughly the average of that of PSA and PSP, the relative influence of the PSA and PSP zones is roughly
neutral with respect to direct simple shear. As a result,
the design method for braced excavation can best be
based on the undrained shear strength of a direct simple
shear test. A similar decision was made by ORourke
(1993).
The equilibrium of the unbalanced weight of soil
inside the mechanism is achieved by mobilization of
shear strength. For each excavation stage, mobilization
A(m,n)
A(m-1,n)
Enlarged strip of the nth layer
C"
D"
Deformation
mechanism of
Excavation stage
m-1
F"
I"
H
n th Layer
Deformation
mechanism of
H" Excavation stage

m
E"
Figure 6. Overlapping of deformation field.
where (m, n) is the total shear strain of the nth layer

in the mth excavation stage, and A(m, n) is the area
of deformation in the nth layer in the mth excavation
stage.
With the help of some suitable stress-strain relation
for the soil (discussed in the following section), the
mobilized strength ratio (m,n) at excavation stage m
for soil layer n can be found (Figure 7).
21
of shear strength of each layer is considered by the

following:
where E and I are the elastic modulus and the second

moment of area per unit length of wall, and s is the
length of the wall in bending. L can be the length of
wall s below the lowest prop.
By assuming the cosine waveform equation (Equation 1), the strain energy term can be shown to be as
follows:
where cu,mob (m,n) is the mobilized undrained shear

strength for layer n in excavation stage m; cu (n) is the
undrained shear strength for layer n; and (m,n) is the
mobilized strength ratio for excavation stage m and
soil layer n.
5.3 Incremental energy balance
where l is the wavelength of deformation, dwmax is the

maximum deflection of the wall in each excavation
increment.
By conservation of energy, the total loss of potential

energy of the soil (P) must balance the total dissipated energy due to plastic shearing of the soil (D)
and the total stored elastic strain energy in bending the
wall (U).
5.4 Calculation procedure

The following calculation procedure is programmed in
Matlab 2006b.
The potential energy loss on the active side of the

wall and the potential energy gain of soil on the passive
side can be estimated easily. The net change of gravitational potential energy (P) is given by the sum of
the potential energy changes in each layer:
1 At each stage of excavation, a maximum deformation wmax , which is bounded by an upper and a lower
bound, is assumed. The soil stratum is divided into
n layers. The areas on both the active side and the
passive side in each layer are calculated.
2 For each layer, with the help of the numerical integration procedure in Matlab, the mobilized shear
strain and the change in PE on both active and
passive sides in different zones is calculated. (Equation 18) The total mobilized shear strain is updated
according Equation 15.
3 With the use of a suitable stress-strain curve (Figure 7), the mobilizable strength ratio can be
found.
4 Total change in PE and total energy dissipation and
elastic bending energy in the wall can be calculated
by Equations 18, 19 & 21, respectively.
5 By considering the conservation of energy of a
structure in statical equilibrium, the sum of total
energy dissipation and elastic strain energy in the
wall balances the total change in PE. To facilitate
solving the solution, an error term is introduced as
follows:
where dwy (m, i) is the vertical component of displacement of soil in the ith layer for the mth construction;
sat (m, i) is the saturated unit weight of soil in the ith
layer for the mth construction.
Since there are no displacement discontinuities, the
total plastic work done by shearing of soil is given by
summing the internal dissipation in each layer:
where cu (m,i) is the undrained shear strength of soil

in the ith layer for the mth construction; d(m,i) is
the shear strain increment of soil in the ith layer for
the mth construction; and the corresponding mobilized
strength ratio is given by:
6 When the error is smaller than a specified convergence limit, the assumed deformation is accepted
as the solution; otherwise, the method of bisection
is employed to assume another maximum displacement and the error term is calculated again using
steps 1 to 5.
7 Then, the incremental wall movement profile is
plotted using the cosine function of equation
8 The cumulative displacement profile is obtained by
accumulating the incremental movement profiles.
The total elastic strain energy stored in the wall, U,

can be evaluated by repeatedly updating the deflected
shape of the wall. It is necessary to do this since U is
a quadratic function of displacement:
22
6 VALIDATION BY NUMERICAL ANALYSIS
Retaining wall
(EI=1440,77 and 19MNm2/m)
s=2.5m
The finite element method can provide a framework

for performing numerical simulations to validate the
extended MSD method in evaluating the performance
of braced excavations. However, finite element analysis of retaining walls is potentially problematic. One
the most difficult problems is the constitutive model
used for the soil. The stress-strain relationship can be
very complicated when considering stress history and
anisotropy of soil (Whittle, 1993).
The validation of the extended MSD method is
examined by comparing its predictions with results of
comprehensive FE analyses of a plane strain braced
excavation in Boston Blue Clay carried out by Jen
(1998). In these analyses, the MIT-E3 constitutive
model is used (Whittle, 1987). The model is based on
Modified Cam clay (Roscoe and Burland 1968). However, several modifications had been made to improve
the basic critical state framework. The model can simulate small strain non-linearity, soil anisotropy and the
hysteretic behaviour associated with reversal of load
paths. Whittle (1993) also demonstrated the ability of
the model to accurately represent the behaviour of different clays when subjected to a variety of loading
paths.
Jen (1998) extended the use of the MIT-E3 model
for analyzing cases of deep excavation in a great
variety of situations. She considered the effect of excavation geometry such as wall length, excavation width
and depth of bed rock, the effect of soil profile such as
cu /OCR ratio and layered soil, and the effect of structural stiffness such as wall stiffness and strut stiffness.
This provides a valuable database for validation of the
extended MSD method.
h=2.5m
L=20, 25, 30, 35 and 40m
C=37.5, 50 and 100m
B/2=15,20,25 & 30m
L
OCR=1
BBC
properties
Figure 8. Scope of parametric study to examine excavation

width effect.
6.1 An example of MSD calculation

The following example shows the extended MSD calculation of wall deflections for a 40 m wall retaining
17.5 m deep and 40 m wide excavation (Figure 8). The
construction sequence comprises the following steps:
Figure 9. Stress-strain response for Ko consolidated

undrained DSS tests on Boston blue clay (After Whittle,
1993).
1 The soil is excavated initially to an unsupported

depth (h) of 2.5 m.
2 The first support is installed at the ground surface.
3 The second level of props is installed at a vertical
spacing of 2.5 m, and 2.5 m of soil is excavated.
MSD method as described above. The total wall movements are then obtained by adding the bulging movements to the cantilever movements to the cantilever
movement according to Clough et al. (1989).
The undrained shear strength of the soil is expressed

by the relationship suggested by Hashash and Whittle
(1996) for Boston Blue Clay (BBC) as follows:
6.1.1 Cantilever movement

By solving for horizontal force equilibrium and
moment equilibrium about the top of the wall, the
mobilized shear stress (cmob ) is found to be 11.43 kPa.
The mobilized strength ratio is 0.2886. With the help
of direct simple shear stress-strain data for Boston blue
clay by Whittle (1993) (Figure 9), the mobilized strain
is read off for an appropriate preconsolidation pressure p and an appropriate OCR. The mobilized shear
The cantilever mode of deformation and the bulging

movements are calculated separately using the mechanism of Osman & Bolton (2006) and the extended
23
factor in estimating the maximum settlement (Mana

and Clough, 1981).
The scope of the excavation analyses are shown in
Figure 8. In the analyses, the excavation was carried
out in undrained conditions in a deposit of normally
consolidated Boston Blue Clay with depth C taken
to be 100 m. A concrete diaphragm wall of depth
L = 40 m, and thickness 0.9 m, supported by rigid
props spaced at h = 2.5 m, was used for supporting
the simulated excavation. The excavation width varies
from 20 m to 60 m. The wavelength of deformation
is chosen according to the l = s rule, where was
taken to be 1.5 and s is the length of wall below the
lowest prop. Computed results by Jen (1998) are used
for comparison. Full details of the analysis procedures,
assumptions and parameters are given in Jen (1998). In
the following section, only results of wall deformation
will be taken for comparison.
Figure 11(a) and (b) show the wall deflection profile with different excavation widths at an excavation
depth of 17.5 m, as calculated by the extended MSD
method and the MIT-E3 model. Figure 11(a) shows
that the excavation width does not have any effect on
the deflected shape of the wall as calculated by the
extended MSD method. Figure 11(b), simularly, shows
a limited effect on the deflected shape of the wall by the
MIT-E3 model. While the MSD-predicted maximum
wall deflection increases by a factor of 1.5 as the width
is increased from 30 m to 60 m, the MIT-E3 computed
maximum wall deformation increased by a factor of
1.6 with the same increase in excavation width.
Figure 10. Wall deflections from MSD with different excavation depths.
strain (mob ) is found to be 0.2%. By considering the

geometrical relationship, the wall rotation is found to
be 0.1%. The displacement at the top of the wall is
found to be 39 mm.
6.1.2 Bulging movement
The first support is installed at the top of the wall.
The length of the wall below the support is 40 m.
By adopting the iterative calculation procedure, and
using the deformation mechanism for a narrow excavation, the bulging movement at each stage of excavation
can be obtained. Then, the incremental bulging movement profile in each stage is plotted using the cosine
function, using the maximum incremental displacement in each stage together with the corresponding
wavelengths. The total wall movement is obtained
by accumulating cantilever movement and the total
bulging movement. Figure 10 shows the final deformation profile of the accumulated wall movement of
an excavation with a width of 40 m. The maximum wall
deflection at an excavation depth of 17.5 m is 115 mm.
The position of the maximum wall displacement is
located at 0.75 L, where L is the length of the wall.
6.3 Effect of bending stiffness of the wall

In general, structural support to excavations is provided by a wall and bracing system. Soldier piles and
lagging, sheet piles, soil mix and soldier piles, drilled
piers (secant piles), and reinforced concrete diaphragm
walls are examples of wall types that have been used to
support excavations. The various types of wall exhibit
a significant range of bending stiffness and allowable
moment. Support walls composed of soldier piles and
sheet piles are generally more flexible and capable of
sustaining smaller loads than the more rigid drilled
piers and reinforced diaphragm walls.
The preceding sections have all assumed a 0.9 m
thick concrete diaphragm wall with elastic bending
stiffness EI = 1440 MNm2 /m. Although it is possible to increase this bending stiffness by increasing
the wall thickness and reinforcement, or by using
T-panels (barettes), most of the walls used in practice
have lower bending stiffnesses. For example, the typical bending stiffness of sheet pile walls is in the range
of 50 to 80 MNm2 /m. This section assesses the effect
of wall bending stiffness on the excavation-induced
displacements.
6.2 Effect of excavation width

The effect of excavation width on predicted ground
movements is the focus of this section. Underground
transportation systems may have excavation widths
ranging from 25 m (a subway station) to 60 m (an
underground highway). The most widely used design
charts generally incorporate the effect of excavation
width in estimation of factor of safety against base
heave (Bjerrum and Eide, 1956) or as a multiplication
24
Figure 12. Deflection profiles of walls with various bending

stiffnesses.
wall stiffness also affect the transition from a subgrade bending mode to a toe kicking-out mode. As the
wall stiffness decreases, the influence of embedment
depth reduces, and hence the tendency for toe kickout to occur is less. Again, a fairly good agreement
can be seen when comparing extended MSD results
and numerical results by the MIT-E3 model, though
kinks are usually found at the wall toe in the numerical predictions, which implies localization of large
shear strains developed beneath the wall toe.
Figure 11. Wall deflection profile of different excavation

widths at H = 17.5 m.
Excavation in soft clay with a width of 40 m supported by a wall of length 25 m and of various bending
stiffness (EI = 1440, 70 and 20 MNm2 /m) are studied.
Results generated by the MSD method and FEA are
compared. Figure 12 (a) and (b) presents the deflection
profiles of the excavations predicted by extended MSD
and the MIT-E3 model, respectively. As the bending
stiffness of the wall decreases, there is no pronounced
change in the overall shape of the wall; the maximum
wall deflection increases and its location migrates
towards the excavated grade. At H = 12.5 m, the maximum wall displacement is 47 mm for the concrete
diaphragm wall with the maximum deflection located
at 7.5 m below the excavation level, while the result
for the most flexible sheet pile wall shows 197 mm
of maximum wall deflection occurring at 5.5 m below
the excavation level. In additional to this, changes in
6.4 Effect of wall length

Wall length is one of the geometrical factors affecting the behaviour of a supported excavation. Previous
analyses were done by Osman and Bolton (2006). The
studies showed that the wall end condition should be
assumed to be free for short walls (L = 12.5 m) since
the clay is very soft at the base and the embedded length
is not long enough to restrain the movement at the tip
of the wall (kick-out mode of deformation). For long
walls (L = 40 m), the embedded depth was assumed
25
Figure 14. Variation of maximum wall deflection with

width to length ratio of wall.
Figure 13. Wall deflection profile of excavation with different support wall lengths, by Extended MSD method.
to be sufficient to restrain the movement at the wall

base (bulging model of deformation). However, the
effect of structural stiffness was not considered in the
old MSD method, though similar observations were
made by Hashash and Whittle (1996) in their numerical
analyses.
In this section, the effect of wall length will be considered. Excavations with widths of 40 m supported by
a 0.9 m thick concrete diaphragm wall with varying
length (L = 20, 25, 30 and 40 m) are studied. Figure 13 shows the wall displacement profiles against
depth with different lengths of wall. For H 7.5 m,
the deflected wall shapes are virtually identical for all
four wall cases of wall length. This agrees with the
conclusion made by Hashash (1992) that wall length
had a minimal effect on pre-failure deformations. As H
increases to 10 m, the toe of the 20 m long wall begins
to kick out with maximum incremental deformations
occurring at the toe of the wall. The movements of
the 25, 30 and 40 m long walls are quite similar. At
H = 12.5 m, the toe of the 20 m and 25 m long wall
kick out, while the two longer walls (L = 30 and 40 m)
continue to deform in a bulging mode. The difference in deformation mode shape demonstrates that the
wall length has a significant influence on the failure
mechanism for a braced excavation.
Figure 15 shows a similar set of analyses by using
the MIT-E3 model. Similar observations about the wall
shape can be made.
Figure 14 summarizes the variation of the normalized excavation-induced deflection (wmax /H ) with
the width to length ratio (B/L) for different bending
stiffnesses of the support wall, for H = 17.5 m.
Figure 15. Wall deflection profile of excavation with different support wall lengths, by MIT-E3 method (After Jen
(1998)).
For a flexible wall (EI = 12.3 MN m2 /m), the normalized maximum wall deflection increases linearly
as the B/L ratio increases from 1 to 2. The gradient
changes and wmax /H increases in a gentle fashion as
the B/L ratio increases from 2 to 2.8. For a rigidly
supported wall, the increase in wmax /H ratio is less significant as the B/L ratio increases. In other words, the
maximum wall deflection is less sensitive to a change
of B/L ratio for a rigid wall.
26
wall. For situations where the wavelength of deformation is restricted less by excavation width than by the
depth of the firm stratum (B > C), the magnitude of
maximum wall deflection increases with the depth of
the firm stratum (C). The MSD method predicts that
the kick-out displacement of the wall toe is limited
by the restriction of developing a large deformation
mechanism. As a result, the maximum wall deflection is also limited. The increase in incremental wall
deflection decreases in later stages of excavation when
H increases from 12.5 m to 17.5 m due to the reduction of wavelength of deformation. On the other hand,
when the depth of the firm stratum is much larger than
the width of the excavation (B < C), the depth of the
bed rock has a minimal effect on the magnitude of
wall deflection. Results by FEA by Jen (1998) (Figure 16(b)) also showed the same observation. Despite
the shortcoming of MSD not being able to model the
correct shape of wall, the maximum wall deflection is
predicted reasonably well. The net difference in maximum wall displacement between MSD and the full
FEA is generally less than 20%.
An extended MSD method is introduced to calculate

the maximum wall displacement profile of a multipropped wall retaining an excavation in soft clay. As
with the earlier MSD approach, each increment of wall
bulging generated by excavation of soil beneath the
current lowest level of support is approximated by a
cosine function. The soil is divided into layers in each
of which the average shear strain increments are compounded so that the mobilized strength ratio in each
layer can be tracked separately as excavation proceeds,
using stress-strain data from a representative element
test matched to the soil properties at mid-depth of the
wall. The incremental loss in potential energy associated with the formation of a settlement trough, due to
wall deformation and base heave, can be expressed as
a function of those ground movements at any stage.
By conservation of energy, this must always balance
the incremental dissipation of energy through shearing
and the incremental storage of elastic energy in bending the wall. By an iterative procedure, the developing
profile of wall displacements can be found.
A reasonable agreement is found between predictions made using this extended MSD method and the
FEA results of Jen (1998) who created full numerical
solutions using the MIT-E3 soil model. In particular,
the effects of excavation width, wall bending stiffness,
wall length, and the depth of the clay stratum, were all
quite closely reproduced.
It is important to draw the right lessons from this.
The excellent work at MIT over many years, on soil
element testing, soil constitutive models, and Finite
Figure 16. Wall deflection profiles of excavation with

different depths to the firm stratum.
6.5
CONCLUDING REMARKS
Effect of the depth of bearing stratum
The depth to bedrock, C, is an important component

of the excavation geometry. The preceding analyses
have assumed a deep clay layer with bedrock located
at C = 100 m which represents a practical upper limit
on C. In practice, however, the clay layer is usually
less than 100 m deep. The following results focus on
the discussion of the geometrical parameter C. The
analysis involves plane strain excavation in normally
consolidated Boston blue clay supported by a 0.9 m
thick concrete diaphragm wall with rigid strut supports
spaced at 2.5 m.
The wall deflection profiles for excavations predicted by both MSD and MIT-E3 with two depths of
the clay stratum (C = 35 m and 50 m) are compared in
Figure 16(a) and (b).
In general, the depth of the firm stratum would only
affect wall deformations below the excavated grade,
hence the largest effects can be seen at the toe of the
27
Bolton, M.D., Powrie, W. & Symons, I.F. 1990b. The design of

stiff in-situ walls retaining over-consolidated clay, part II,
long term behaviour. Ground engineering 23(2): 2228.
Clough, G.W. & Hansen, L.A. 1981. Clay anisotropy and
braced wall behavior. ASCE Journal of Geotechnical
Engineering 107(7): 893913.
Clough, G.W., Smith, E.W. & Sweeney, B.P. 1989. Movement
control of excavation support system by iterative design.
Foundation Engineering Current Principles and Practices
Vol. 2 ASCE, New York, NY: 869882.
Eide, O., Aas, G. & Josang, T. 1972. Special application of
cast-inplace walls for tunnels in soft clay. Proceedings
of the 5th European Conference on Soil Mechanics and
Foundation Engineering, Madrid, Spain, 1: 485498.
Hashash, Y.M.A. & Whittle, A.J. 1996. Ground movement
prediction for deep excavations in soft clay. ASCE Journal
of Geotechnical Engineering 122(6): 474486.
Jen, L.C. 1998. The design and performance of deep excavations in clay. PhD thesis, Dept. of Civil and Environmental
Engineering, MIT, Cambridge, Mass.
Mana, A.I. & Clough, G.W. 1981. Prediction of movements
for braced cut in clay. J. Geo. Engrg. ASCE 107(GT8):
759777.
Milligan, G.W.E. & Bransby, P.L. 1976. Combined active
and passive rotational failure of a retaining wall in sand.
Geotechnique 26(3): 473494.
ORourke, T.D. 1993. Base stability and ground movement prediction for excavations in soft clay. Retaining
Structures, Thomas Telford, London: 131139.
Osman, A.S. & Bolton, M.D. 2004 A new design method for
retaining walls in clay. Canadian Geotechnical Journal
41(3): 451466.
Osman, A.S. & Bolton, M.D. 2005 Simple plasticity-based
prediction of the undrained settlement of shallow circular
foundations on clay. Geotechnique 55(6): 435447.
Osman, A.S. & Bolton, M.D. 2006 Ground movement predictions for braced excavations in undrained clay. ASCE Journal of Geotechnical and Geo-environmental Engineering
132(4): 465477.
Osman, A.S., Bolton, M.D. & Mair, R.J. 2006 Predicting
2D ground movements around tunnels in undrained clay.
Skempton, A.W. 1951. The bearing capacity of clays. Proc.,
Building Research Congress, London: 180189.
Sloan, S.W. 1988. Lower bound limit analysis using finite
elements and linear programming. International Journal
for Numerical and Analytical Methods in Geomechanics
12(1): 6177.
Sloan, S.W. & Kleeman, P.W. 1995. Upper bound limit
analysis using discontinuous velocity fields. Computer
Methods in Applied Mechanics and Engineering 127:
293314.
Terzaghi, K.1943. Theoretical soil mechanics, John Wiley &
Sons, Inc., New York, N.Y.
Ukritchon, B., Whittle, A.J. & Sloan, S.W. 2003. Undrained
stability of braced excavations in clay. ASCE Journal of
Geotechnical and Geoenvironmental Engineering 129(8):
738755.
Whittle, A.J. 1987. A constitutive model for overconsolidated
clays with application to the cyclic loading of friction
piles. PhD thesis, Dept. of Civil and Environmental
Engineering, MIT, Cambridge, Mass.
Whittle, A.J. 1993. Evaluation of a constitutive model for
overconsolidated clays. Geotechnique 43(2): 289313.
Element Analysis, have provided us with the means to

calibrate a very simple MSD prediction method. This
was based on an undrained strength profile, a single
stress-strain test, and a plastic deformation mechanism. Were it not for the multiple level of props the
calculation of ground displacements could be carried
out in hardly more time than is currently required to
calculate a stability number or factor of safety. Allowing for the need to represent various levels of props, the
calculations then call for a Matlab script or a spreadsheet, and the whole process might take half a day to
complete.
An engineer can therefore anticipate that important
questions will be capable of approximate but reasonably robust answers in a sensible industrial timescale.
For example:
Will a prop spacing of 3m be sufficient for a wall
of limited stiffness and strength?
Should the base of the wall be fixed by jet-grouting
prior to excavation?
Will a particular construction sequence cause the
soil to strain so much that it indulges in post-peak
softening?
Is it feasible to prop the wall at sufficiently close
spacings to restrict strains in the retained ground to
values that will prevent damage to buried services?
This may lead engineers to take soil stiffness more
seriously, and to request accurate stress-strain data. If
so, in a decade perhaps, our Codes of Practice might
be updated to note that MSD for deep excavations provides a practical way of checking for the avoidance of
serviceability limit states.
ACKNOWLEDGEMENT
The authors would like to acknowledge the earmarked
research grant # 618006 provided by the Research
Grants Council of the HKSAR Government, and
also the Platform Grant (GR/T18660/01) awarded by
the UK Engineering and Physical Sciences Research
Council.
REFERENCES
Bjerrum, L. & Eide, O. 1956. Stability of strutted excavations
in clay. Geotechnique 6: 115128.
Bolton, M.D. & Powrie, W., 1988. Behaviour of diaphragm
walls retaining walls prior to collapse. Geotechnique
37(3): 335353.
Bolton, M.D., Powrie W. & Symons, I.F. 1989. The design
of stiff in-situ walls retaining overconsolidated clay part.
Ground engineering 22(8): 4448.
Bolton, M.D., Powrie, W. & Symons, I.F. 1990a. The design of
stiff in-situ walls retaining over-consolidated clay part 1,
short term behaviour. Ground Engineering 23(1): 3439.
28
Overview of Shanghai Yangtze River Tunnel Project

R. Huang
Commanding Post of Shanghai Tunnel & Bridge Construction, Shanghai, P.R. China
ABSTRACT: In the paper, an introduction of the construction background and scale of ShanghaiYangtze River
Tunnel and Bridge Project and natural conditions of Shanghai Yangtze River Tunnel construction are given. The
overall design concept and some critical technical solutions such as segment structure of large diameter bored
tunnel, water proofing of segment under high depth and water pressure, long tunnel ventilation system and fire
fighting system are described. Characteristics of two mixed TBM with a diameter of 15,430 mm are described.
The overall construction methods of tunnel, and critical technical solutions and risk provision measures for large
and long river-crossing tunnel such as the front surface stability for bored tunnel construction, floating resistance
of large diameter tunnel, long distance construction survey, synchronous construction of internal structure, and
cross passage construction of fresh/salty alternating geological/environmental condition are discussed.
INTRODUCTION
Shanghai Yangtze River Tunnel and Bridge project is

located at the South Channel waterway and North
Channel waterway of Yangtze River mouth in the
northeast of Shanghai, which is a significant part
of national expressway, as shown in Figure 1. It is
an extremely major transport infrastructure project at
seashore area in China at Yangtze River mouth and
also the largest tunnel and bridge combination project
worldwide. The completion of the project will further
promote the development space for Shanghai, improve
the structure and layout of Shanghai traffic system,
develop resources on Chongming Island, accelerate economic development in the north of Jiangsu
Province, increase the economy capacity of Pudong,
accelerate the economy integrity of Yangtze River
Delta, boom the economic development of Yangtze
River area and even the whole country and upgrade
the comprehensive competence of Shanghai in China
and even in the global economy.
ShanghaiYangtze River Tunnel and Bridge (Chongming Crossing) alignment solution is the planned
western solution which is implemented firstly based on
the Shanghai overall urban planning, and comparison
between east and west alignment and in combination
of various aspects. The western alignment starts from
Wuhaogou in Pudong, crossing Yangtze River South
Channel waterway to Changxing Island and spanning
Yangtze River North Channel waterway to east of
Chongming Island.
Yangtze River begins to be divided into 3 levels of
branches and have 4 mouths flowing into the sea: The
Figure 1. Site location of Chongming Crossing.
South Channel waterway is mixed river trench. The

intermediate slow flow area forms Ruifeng shoal
which is relatively stable for a long time. The natural
water depth makes it as the main navigation channel.
However, the North Channel waterway is located in
the middle part of river, which is influenced by the
south part and branch transition into North Channel
waterway. So the trench varies alternatively and the
river map is not as stable as South Channel waterway.
Therefore, after iterative discussion by several parties,
finally the solution of Southern Tunnel & Northern
Bridge is selected. The total project is 25.5 km long,
among which 8.95 km is tunnel with a design speed
of 80 km/h and 9.97 km is bridge and 6.58 km is land
connection with a design speed of 100 km/h, as shown
29
Figure 2. Diagram of Shanghai Yangtze River Tunnel and Bridge.
River Tunnel and Bridge Construction Development

Co., Ltd. which is mainly in charge of the implementation of the project and daily work of commanding
post and performs the investment management on
behalf of the client. The specific work is responsible
for the financing, investment, construction, operation
and transfer of the project. To detail the technical
assurance measures, the clients sets up the technical
consultant team which provides theoretical support,
technical assistance and consultancy service for significant technical challenges during the implementation.
Meanwhile, the team is involved in the investigation of
significant technical solutions, review of construction
method statement and treatment of technical problems
to ensure the high quality and safety. International
well-known consultancy companies are entrusted for
the purpose of application of state-of-art philosophy,
most successful experience, optimal concept and most
mature management to make theYangtze River Tunnel
and Bridge Project as Century Elite Project.
The project finally initiated on 28th, December
2004 and planned to be open to traffic in July 2010.
The main civil structure of the bridge is planned to be
closed in June 2008, and tunnel in April 2009.
in Figure 2. The total roadway is planned as dual

6 lanes.
2
CONSTRUCTION BACKGROUND AND

PLANNING
The planning study of Shanghai Yangtze River Tunnel and Bridge Project (Chongming Crossing) was
incepted from 90s of last century. The preliminary
preparatory work has lasted 11 years. In May 1993, the
National Scientific Committee held the Yangtze River
mouth crossing significant technical-economical challenges early stage work meeting. After one year special investigation, the Preliminary study report of significant technical challenges of Yangtze River Crossing was prepared. The pre-feasibility study report was
prepared in March 1999. In August 2001, the international concept competition was developed and the
Southern Tunnel & Northern Bridge solution was
defined. The National Planning Committee approved
the project proposal in December 2002. The feasibility
study report was approved by the National Development and Reform Committee in November 2004. The
preliminary design was approved by the Ministry of
Communication in July 2005 and total investment of
12.616 billion RMB was approved for the project.
For the project construction investment, 5 billion
was funded by Shanghai Chengtou Corporation (60%)
and Shanghai Road Construction Cooperation (40%),
and 7.6 billion was financed from Bank Consortium.
Based on the characteristics of the national major
project, Commanding Post of Shanghai Tunnel &
Bridge Construction was established with approval of
Shanghai Municipal Committee. The post is directed
by the vice major and composed of staff from Pudong
New Area, Chongming County and other committees
and bureaus. The main responsibility is to make decision on significant problems and coordinate important
items. In order to improve the depth of daily management, office was set up under the commanding post,
working together with established Shanghai Yangtze
NATURAL CONDITIONS OF TUNNEL

PROJECT
3.1 Environmental conditions

Shanghai Yangtze Tunnel Project starts from Wuhaogou of Waigaoqiao in Pudong New Area, connected
with Shanghai main fast roads such as Middle Ring,
Outer Ring and Suburb Ring through Wuzhou Aveneu,
crossing southern water area and lands on Changxing Island 400 m west of Xinkaihe Harbour, connected
with Changxing Island road net through Panyuan Interchange. The main building on land is the flood prevention wall on Pudong side and Changxing Island. Others
are farm fields.The river-crossing section is mainly the
30
Figure 3. Longitudinal profile of tunnel.
hydraulic relation with river water. The potential water

level is mainly influenced by the Yangtze River flux
and ebb. The average water level for Waigaoqiao and
Changxing Island is 2.8 m and 2.4 m, respectively.
In the stratum and at site area, the confined water is rich. At most area, the confined water
is directly continuous. The confined water level is
between 4.15 m and 6.76 m. Furthermore, slight
confined water distributes in 2 , which has certain
hydraulic relations with confined water in .
southern water way for navigation which is an important passage for connecting Yangtze Waters with other
seashore area in China and oceans worldwide.
There are two sea cables arranged along the bored
tunnel axis with a depth of 3 m below natural river
bed. One cable is basically located at the west side
of the tunnel and goes into the river near Wuhaogou
on Pudong side, which is about 1,500 m away from
the tunnel. It becomes closer to the tunnel gradually
to the north and crosses the tunnel to its east at 240 m
from Changxing Island and lands on Changxing Island
at 350 m west of Xinkaihe Harbour. The other cable
goes into the river near Wuhaogou, 1,300 away from
the tunnel. Then it turns to NE first and N at 2,600 m
way from Pudong Land Connections, almost identical
with the tunnel alignment. And it changes from the
west of tunnel to east of tunnel gradually and lands
on Changxing Island about 300 m west of Xinkaihe
Harbor.
Furthermore, two sunken boats close to Chainage
XK2+350 and XK1+500 have been salved before
bored tunnel construction. Earth was also filled back
at corresponding locations; however, there may be still
some remains.
3.2
3.3
Geological conditions
The relief of onshore area of the project is river mouth,

sand mouth, sand islandwhich is within the major four
relief units in Shanghai. The ground surface is even
with a normal elevation of 3.5 m (Wusong Elevation).
The water area is classified as river bed relief.
The project site has a seismic fortification intensity
of 7, classified as IV site. The stratum 3 and 2
sandy silt distributing on Pudong land area is slightly
liquefied.
Main geological layers (refers to Figure 3) TBM
crosses are: 1 grey muddy clay, 1 grey muddy
clay, 2 grey clayey silt with thin silty clay, 3 silty
clay, 3 tlens, 11 grey clay silt, 12 grey sandy
silt, etc. Unfavorable geological conditions are experienced along the axis of the tunnel, such as liquefied
soil, quick sand, piping, shallow gas (methane), lens
and confined water, etc.
River regime and hydrological conditions
At the mouth of Yangtze River it is tide area with intermediate level. Outside of mouth is regular half day tide
and inside is irregular half day shallow tide due to the
change of tide wave. Average flood tide time is 5 h and
average ebb tide time is 7 h, so total time for ebb and
flux is 12 h. The average currency flow is 1.05 m/s for
flood tide during flood season and 1.12 m/s for ebb
tide. The maximal flow for flood tide is 1.98 m/s and
2.35 m/s for ebb tide.
The underground water type in the shallow stratum at tunnel site is potential water, which has close
4 TUNNEL DESIGN SOLUTION

4.1
Scale
Shanghai Yangtze River Tunnel is designed as dual

6 lanes expressway, and rail traffic provision is made
below the road deck. Seismic fortification level is 7.
31
Figure 4. Cross section of bored tunnel.
for rail traffic provision and power cable gallery and

4 is for waste water pump plant.
The cut-and-cover is designed with a rectangular
shape consisting of two tubes and one cable channel. 3 lanes are arranged in each tube. The structural
limit is 13.25 m in width and 5.5 m in height, as shown
in Figure 5. Upper area with a height of 0.6 m is for
equipment provision. The upper part of central gallery
is for cable channel, middle part for evacuation and
lower part for pipe ditch. Ventilation shaft and building
for equipments are arranged above the cut-and-cover
tunnel close to the working shaft.
The approach consists of light transition zone and
open ramp. The structural limit of cross section is identical with that of cut-and-cover tunnel. Both sides have
a slope section with a slope of 1:3 with green planting
for protection. The light transition zone is designed as
steel arch structure.
Design service life is 100 years. The project consists of land connections of Pudong side (657.73 m),
river-crossing tunnel (east tube 7,471.654 m and west
tube 7,469.363 m) and land connections on Changxing Island (826.93 m). Total length is 8,955.26 m and
investment is 6.3 billion RMB. The river-crossing part
is twin-tube bored tunnel.
4.2 Tunnel alignment
The longitudinal profile of bored tunnel is in a shape of
W with a longitudinal slope of 0.3% and 0.87%. The
land connections have a longitudinal profile of 2.9%.
The minimal curvature radius of horizontal plane is
4,000 m and vertical profile 12,000 m.
4.3
Building design
4.3.1 Cross section of bored tunnel

Based on structural limit of traffic passage and equipment layout requirement, the internal diameter of
lining for bored tunnel is determinated as 13.7 m
considering the fitted tolerance of lining at curved
section, construction tolerance, differential settlement,
and combining the design and construction experience. On the top of tunnel, smoke discharge ducts are
arranged for fire accident with an area of 12.4 m2 . Each
tunnel has three lanes with a structural clear width of
12.75 m and road lane clear height of 5.2 m. The central part below road deck is for rail traffic provision in
future. On the left side, beside the buried transformer
arrangement, it also serves as main evacuation stairs.
The right side is cable channel, including provision
space for 220 kV power cable, as shown in Figure 4.
4.4 Structural design

4.4.1 Structural design of bored tunnel
The external diameter of bored tunnel lining is
15,000 mm and internal diameter 13,700 mm, as
shown in Figure 6. The ring width is 2,000 mm and
thickness is 650 mm. Precast reinforced concrete common tapered segments are assembled with staggered
joint. Concrete strength class is C60 and seepage resistance class is S12. The lining ring consists of 10
segments, i.e. 7 standard segments (B), 2 adjacent
segments (L), and 1 key segment (F). According to
the different depth, segments are classified as shallow segments, middle-deep segments, deep segments
and extremely deep segments. Skew bolts are used to
connect segments in longitudinal and circumferential
direction. 38 M30 longitudinal bolts are used to connect the rings. 2 M39 circumferential bolts are used
to connect the segments. Shear pins are added between
4.3.2 Cross-section of land connections

Working shaft is underground four-floor building:
1 is for ventilation pipe and pump plant for fire
fighting; 2 is for road lane with cross over; 3 is
32
Figure 5. Cross-section of cut-and-cover.
Figure 6. Lining structure.
For the open cut ramp, the bottom plate structure

thickness is around 5001,100 mm. Under the bottom
plate, bored piles are arranged as up-lifting resistance
pile to fulfil the structural floating resistance requirement. The slope uses in-situ cast reinforced concrete
grid and fill earth and green planting in the grid for
protection.
lining rings at shallow cover area, geological condition

variation area and cross passage to increase the shear
strength between rings at special location and reduce
the step between rings.
4.4.2 Structural design of land connections
The working shaft and cut & cover tunnel share the
same wall. The thickness of diaphragm of working
shaft is 1,000 mm, and the inner wall is 500 mm,
1,200 mm, respectively. For the cut-and-cover tunnel,
the thickness of diaphragm is 1,000 mm, 800 mm,
and 600 mm respectively depending on the excavation
depth. The inner structure thickness is 600 mm.
4.5
Structural water-proof and durability design
4.5.1 Requirement and standard

For the bored tunnel and working shaft, the water proof
standard of slightly higher than level II is required. For
33
(a)
(b)
Figure 7. Segment joint water proofing sketch.
4.6 Tunnel operation system
the entire tunnel, the average leakage should be less

than 0.05 L/m2 d. For each random 100 m2 , the leakage should be less than 0.1 L/ m2 d. The inner surface
wet spots should not be more than 4 of total inner
specific surface area. In each random 100 m2 , the wet
spots should not be more than 4 locations. The maximal area of individual wet spot should not be large
than 0.15 m2 .
The chloride diffusion coefficient of concrete lining structure of bored tunnel is not more than
12 1013 m2 /s. Concrete seepage resistance class is
not less than S12. Furthermore, it is required that under
1 MPa water pressure which is equivalent to 2 times of
water pressure for the tunnel with the largest depth, no
leakage is occurred when the lining joint opens 7 mm
and staggers 10 mm. The safety service life of water
proof material is 100 years.
The seepage resistance class of onshore tunnel
structure is not less than S10.
4.6.1 Ventilation system

The road tunnel uses jet fan induced longitudinal
ventilation combined with smoke ventilation.
The longitudinal ventilation area in tunnel is 82 m2 .
Jet fans are suspended above the deck lane and below
the smoke discharge duct, supporting induced ventilation in normal operation and congested condition.
78 jet fans with a diameter of 1,000 mm are arranged
in each tube from Pudong access to Changxing Island
access, every 3 as a group.
Ventilation shafts are arranged on Pudong side and
Changxing Island, respectively, housing large ventilation machine and special smoke discharge axial fan.
The fans are connected with main tunnel through air
inlet and ventilation duct. During normal operation
and congested condition, the ventilation machine is
turned on to discharge the polluted air in the tunnel. 6
large axial fans with a capacity of 75 m3 /s 150 m3 /s
are housed in the working shaft on Changxing Insland
and Pudong, respectively.
For normal operation of lower rail traffic, piston
ventilation mode is used.
4.5.2 Water proofing design

The segment joint water proof arrangement consists of
EPDM rubber strip with small compressive permanent
deformation, small stress relaxation and good aging
resistance performance and hydrophilic rubber strip,
as shown in Figure 7.
The deformation joint at cut-and-cover tunnel uses
embedded water stop gasket, outer paste gasket and
inserted sealing glue forming enclosed system. The
top plate uses water proof paint as outer water
proof layer.
4.6.2 Water supply and drainage system

The fire water, washing waste water, and structural
leakage are collected by the waste water sump at the
lowest point of river. Sump is arranged at upper and
lower level, respectively. The lower waste water is
drained by the relay of upper sump. The upper sump
is arranged on two sides of rail traffic area, housing
34
four pumps which are used alternatively under normal

operation and turned on entirely during fire fighting. For lower level, 4 sumps with a dimension of
1,000 1,000 550 mm are arranged at the lowest
point of tunnel where SGI segment is used and above
the sump water collection trench with a length of 7 m
and a width of 1 m is arranged. One waste water pump
is placed in each pit which are used alternatively at
normal condition and three are used, one spare during
fire fighting.
At each access of tunnel, one rain water sump is
arranged to stop water and drain it out of the tunnel.
The rain amount is designed based on a return period
of 30 years for rainstorm.
monitoring system, and expressway net traffic monitoring emergency center, rail traffic monitoring and
220 kV, etc.
The information collected by the tunnel monitoring system, bridge monitoring system, and toll station
system is transferred to the monitoring and control
center in the tunnel and bridge administration center
on Changxing Island. Furthermore, one administration center is arranged at Wuhaogou on Pudong side
assisting the daily management and emergency treatment, establishing the three level frame of monitoring
and control center administration center outfield
equipment.
4.6.3 Power supply system

The electricity load in tunnel is classified as three levels: level I is for ventilation fan, valve, water pump,
lighting and monitoring & control system and direct
current screen, etc; level II is for tunnel inspection and
repair, and ventilation fan in transformer plant; level III
is for air conditioning cold water machines.
On Pudong side and Changxing Island, two transformers are arranged. Two independent 35 kV power
circuits are introduced respectively and can be used
as spare power for the other through two connection cables. Each route ensures the electricity load of
level I and II in the tunnel. For the dynamical and
lighting load far away from transformers, the power
is supplied through 10 kV power supply network in
the tunnel and embedded transformers underneath the
tunnel to ensure the long distance power supply quality and reduce energy losses. 6 kV power is supplied
for the concentrated ventilation fan. Lighting electricity is supplied by independent circuit in power supply
system.
4.7
Fire-fighting system
The fire fighting sytem design cosists of balanced

and redundant design of safety and function for the
entire tunnel structure, building, water supply and
drainage and fire fighting, emergency ventilation and
smoke discharge, lighting, power supply and other
subsystems. The details are as follows:
Cross passage is arranged every 830 m connecting
the upchainage and downchainage tunnel for passenger evacuation with a height of 2.1 m and width
of 1.8 m. Three evacuation ladders are arranged
between two cross passages connecting the upper
and lower level.
The passive fire proof design uses the German
RABT fire accident temperature rising curve. The
fire accident temperature is 1,200 C. Fire proof
inner lining which ensures the surface temperature of protected concrete segment is not more than
250 C within 120 minutes is selected to protect
the arch above smoke duct, smoke duct and crown
above the finishing plate. For rectangular tunnel,
fire proof material which ensures the structure top
plate safety within 90 minutes is selected to protect the top plate and 1.0 m below the top plate. To
ensure the passenger evacuation, fire proof bursting
resistance fibre is mixed in the concrete bulkhead
to achieve no damage of structure when structure
is exposed to fire for 30 minutes.
The ventilation system is designed based on only
one fire accident in road tunnel and rail traffic area.
The marginal arch area of bored tunnel is used
for smoke duct. Special smoke ventilation valve
is arranged every 60 m for the smoke ventilation
in case of fire accidents on road level. When fire
accident occurs in lower level, ventilation fan in the
working shaft is turned on to ventilate the smoke
to the side of fire source while passengers evacuate
towards the fresh air.
The emergency lighting is arranged on two sides
with the same type. As the basic lighting, inserted
into the basic lighting uniformly. Meanwhile, normal lighting and emergency lighting are installed in
4.6.4 Lighting system

Light belt is used for lighting in the tunnel. At portal
area, natural light transition and artificial light combination is used for lighting. Fluorescence lamp is the
main light source in the tunnel. Strengthening lighting
uses the high pressure sodium lamp. Taking account
of the energy consumption, the application research
of LED with high power is being developed. The shift
time for emergency lighting in the tunnel should not
be larger than 0.1 s and the emergency time is 90 min.
4.6.5 Monitoring and control system
The comprehensive monitoring system consists of traffic monitoring system, equipment monitoring system,
CCTV monitoring system, communication system,
fire automatic alarming system, central computer
management system, monitoring and control center.
Equipment monitoring system is classified as ventilation subsystem, water supply and drainage subsystem,
lighting subsystem, and electrical monitoring subsystem. Monitoring system has access provision for health
35
the cable passage. Evacuation guidance signs are

arranged on the two sides of road, cross passage
and safety passage. Emergency telephone guidance
signs are arranged above the telephones in tunnel.
Fire water supply at both ends of tunnel is from
the DN250 water supply pipe introduced from two
different municipal water pipes without fire water
pond. One fire fighting pump plant is arranged
in working shaft on Pudong side and Changxing Island, respectively. The fire hydrant system
is continuous in the longitudinal evacuation passage. Fire hydrant group is arranged every 50 m at
one lane side in each tunnel and fire extinguisher
group every 25 m. Foam-water spraying system is
used in the tunnel which can provided foam liquid
continuously for 20 min and arranged every 25 m.
The communication and linkage of each subsystem of comprehensive monitoring and control
system can realize the monitoring, control and
test of the whole tunnel such as fan, water pump,
electrical and lighting equipment. Fire automatic
alarming system can detect the possible hazards
such as fire fast, real-time identify and alarm and
has the function of passage alarming and tunnel
closed. Furthermore, corresponding equipments
can be automatically activated to extinguish the fire
at early time and organize the hazard prevention to
reduce the loss to the minimum extent.
5
Figure 8. Bulkhead of Mixshield TBM.
degrees. Vacuum suction plated is used to grasp the

segment. 6-point grouting is used for simultaneous
grouting.
Backup system consists of 3 gantries: gantry 1 housing the power equipment and control system, gantry
2 housing 3 cranes and bridge section for segment,
road element, and other construction material transport, gantry 3 is pipe laying gantry for carrying the
extension of the different services such as cable hose,
slurry, air and industrial water pipes.
Excavated soil is transported from excavation
chamber to the slurry treatment plant (STP) through
the slurry pipe in the slurry circulation system. After
separation by the treatment equipment, excavated soil
with large size is separated and then the recycled slurry
is pumped back into excavation chamber and working
chamber.
15, 430 MM SLURRY MIXED TBM
Two large slurry pressurized mixed shield machines

with a diameter of 15.43 m are used for the construction of 7.5 m long bored tunnels.
5.2 Adaptability to the large, long and deep

characteristics
5.1 TBM performance and characteristics
For the TBM construction, firstly the project and crew

safety should be ensured. The key for safety of TBM is
to protect the cutter head and tailskin, mainly including
cutter head design, main bearing seal and tailskin seal
assurance. Furthermore, the maintenance and repair
of these parts are risk and difficult to access, so the
inspection and possibility for maintenance in case of
failure must be considered.
The TBM consists of shield machine and backup system with a total length of 13.4 m and weight of 3,250 t,
including cutter head system, shield body, tailskin,
main drive, erector, synchronized grouting system,
transport system, guidance system and data acquisition
system and slurry system.
The TBM has excavation chamber and working
chamber. During advancing, the air bubble in the working chamber is adjusted through the control unit to
stabilize the slurry level thus balance the water/soil
pressure in excavation chamber, as shown in Figure 8.
The thrust system consists of 19 groups thrust cylinders with a total thrust force of 203,066 kN. Cutter
head is drived by 15 motors with 250 kW power, so
the total power is 3,750 kW. Installation position for
two spare motors is also provided. Tailskin seal structure is composed of three rows steel wire brushes and
one steel plate brush, forming 3 grease chambers. The
erector system is centrally supported with 6 freedom
5.2.1 Cutter head and cutting tools

Cutter head is for soft ground and can be rotated in
two directions. The cutter head is pressure resistant
steel structure and specific wear protection is arranged
for the periphery area. Special wear protection is also
designed for cutting tools.
The closed type cutter head is designed with 6
main arms and 6 auxiliary arms, 12 large material
opening and 12 small material opening. The opening
ratio is around 29%. 209 cutting tools are arranged on
36
Figure 10. Tailskin structure.

Figure 9. Main bearing seal arrangement.
Furthermore, 1 emergency seal is arranged between

the 3rd row steel wire brush and the steel plate brush.
The emergency seal has the function to protect the ring
building area from water ingress while changing the
first three steel brush seals. Due to no practical application references of this technology, modeling test has
been carried out for the emergency seal installation to
confirm the reliability of the emergency seal when the
inflatable seal is pressurized to 1 MPa.
Simultaneous grouting lines are arranged at the
tail skin, including one standard grout pipeline and
one spare pipeline for filling the annulus gap outside the segment after excavation. Furthermore, 19
chemical grout pipes are added for special hardening
grout (simultaneous slurry penetrating into cement)
or polyurethane for leakage block in emergency condition. 19 3 grease pipes have the function of steel
wire brush lubrication and tail skin sealing. The seal
system is controlled from the cabinet in automatic and
manual modes through time and pressure control.
Furthermore, freezing pipelines are arranged at the
tailskin to ease the ground treatment by means of freezing measures in case of leakage and ensure the seal
treatment and repair safety.
the cutter head, among which 124 fixed scrapers, 12

buckets, 2 copy cutters, 7 replaceable center tools and
64 replaceable tools.
The scrapers are custommade large tools with features of 250 mm width, wear-resistance body and
high quality carbide alloy cutting edges whose angle
matches the parameter of excavated ground.The scrapers at the edge are used to remove the excavated soil
at edge and protect the cutter head edge from direct
wear. Copy cutter can automatically extend and retract.
The multiple over-cut areas can be setup in the control cabin and corresponding cutting tools position are
displayed. The replaceable cutting tools have special
seal to prevent the slurry at the front surface enter into
the cutter head chamber. During operation, the workers can enter the cutter head chamber to replace the
cutting tools under atmospheric condition with high
safety, good operation possibility and low risk.
In order to avoid clogging at cutter head center,
the opening at center is designed as chute to ease the
material flowing. Meanwhile, one bentonite hole is
arranged at each opening to ease flushing in case of
clogging.
5.2.2 Main bearing seal
Two sets seal system are arranged for the main bearing seal design. The outer seal is for the excavation
chamber side and inner seal for the shield body with
normal pressure. The special seal combination can
bear a pressure of 8.5 bar.
Outer seal is to separate the main bearing and excavation chamber. Seal type is axial seal with large
diameter, totally 4 lip seals and one pilot labyrinth,
thus forming 4 separate areas, as shown in Figure 9.
The inner seal one the gear box side is special axial
seal which can carry the pressure of gear chamber.
The seal system has grease lubrication and leakage monitoring system which can monitor the grease
amount by pressure and flow monitoring. The seal system has been proved successfully in many projects for
several years and become a standard configuration.
5.2.4 Man lock and submerged wall

During long distance advancing, there is a possibility
of operation failure of mixing machine due to large
obstacles blocking such as stones, main bearing seal
replacement due to wear, submerged wall closed or
leakage examination in the air bubble chamber. These
maintenance and repair work need workers access
the air bubble chamber with a pressure up to 5.5 bar.
Therefore, two man locks are arranged to ensure the
maintenance and repair workers can access.
The main chamber of manlock can house one 1.8 m
stretcher. Under pressure-reducing condition, the medical staff can access the main chamber and organize
rescue in case of emergency. Meanwhile, the other man
lock can transport the tools, material and equipment
from TBM to the air bubble chamber.
The man lock is equipped with poisonous gas detection system which can take the sample of enclosed
gas in the man lock. The system information will
be displayed at the working position where outside
5.2.3 Tailskin
The tailskin is sealed off by 3 rows steel wire brush
and 1 steel plate brush, as shown in Figure 10.
37
staff is. The man lock also provides the flange connection. Once the rescue and injuries enters into
temporary rescue chamber, the temporary chamber
can be disassembled fast and transported out of the
tunnel, connected with large medical chamber for the
convenience of medical work to rescue.
The submerged wall uses hydraulic drive and is
equipped with air pressure seal strip. When normal
operation in the working chamber is needed, the submerged wall can be closed thus the excavation chamber
and working chamber can be separated, and then the
valve can be opened for reducing the pressure. At this
time, pipe for supplementing slurry which penetrates
working chamber can maintain the slurry pressure in
the excavation chamber.
Figure 11. Water stop tank sketch.
beyond the treated ground area and holes are bored for
grouting the annulus to ensure the safety during tunnel gate removal. These three measures application has
achieved good performance. During TBM launching,
the treated soil is stable.
6 TUNNEL CONSTRUCTION METHOD

6.1
(2) Tunnel annulus seal

The diameter of tunnel eye is up to 15,800 mm. To prevent the slurry enters into the working shaft from the
circular build gap between tunnel eye and shield or segment during launching thus affect the establishment of
front face soil and water pressure, good performance
seal water stopping facility is arranged. The facility is
a box structure with 2 layers water stop rubber strip and
chain plate installed, as shown in Figure 11. The outside chain plate is adjustable with 50 mm adjustment
allowance. Furthermore, 12 grout holes are arranged
uniformly along the outside between two layer water
stop on the box for the purpose of sealing in case of
leakage at the tunnel eye. The outer end surface of
water stop facility shall be vertical to the tunnel axis.
Overall arrangement and time schedule
Based on the overall programming, the construction of

working shafts, bored tunnel, synchronous construction of road structure, operation equipment installation
and commissioning are the main works and secondary
works such as receiving shaft and crosspassage in
parallel.
In May 2006, the launching shaft and onshore structures on Pudong side were completed and site assembly
of two TBMs started. The east tunnel starts advancing
in September 2006, while west tunnel in January 2007.
During construction of these two tunnels, the prefabricated road element erection and TBM advancing
are synchronous, which on one hand resist the tunnel floating during construction stage and on the other
hand provide special truck passage for segments, prefabricated road elements and materials to realize the
fast bored tunnel construction. In parallel with bored
tunnel construction, the road deck structure construction is also carried out 200250 m back from segment
erection and top smoke duct will start construction in
January, 2008, forming gradually working flow in tunnel. After west tube TBM advancing 3 km, the first
crosspassage started construction in October, 2007.
After the tunnel is through, final connection work of
working shaft and road structure is carried out and
operation equipment and finishing and pavement work
will start.
6.2
(3) Back support for TBM

The back up shield support includes 7 rings, among
which -6 is steel ring composed of 4 large steel segments with high fabrication quality to ensure the
circularity and stiffness of the reference ring, as shown
in Figure 12. After precise positioning of steel ring, it
is supported on the concrete structure of cut and cover
tunnel by 19 steel struts with a length of 1.2 m. Other
6 minus closed rings segments are assembled with
staggered joint. Inserts are embedded on the inside
and outside surface. After each ring building, the circumferential ring and longitudinal ring are connected
with steel plate to improve the integrate stiffness and
ensure the circularity and ring plane evenness. Meanwhile, the circumferential plane of each minus ring
shall be vertical to the design axis.
Main critical technical issues during bored

tunnel construction
6.2.1.2 TBM receiving

(1) Arrangement in receiving shaft
Before TBM receiving, the diaphragm between receiving shaft and cut & cover tunnel and the diaphragm
in the receiving shaft between upchainage and downchainage tunnel shall be completed to make the receiving shaft as an enclosed shaft structure. Then MU5
cement mortar is cast in the working shaft with a height
6.2.1 TBM launching and arriving technology

6.2.1.1 TBM launching
(1) Tunnel eye stabilization
3-axial mixing pile and RJP injection procedure is
used surrounding the working shaft to stabilize the
ground forming a stabilized area of 15 m in length.
6 dewatering wells for bearing water are supplemented
38
Figure 12. Back supports for TBM.
could be released slowly under close observation. If

any water leakage is observed, the polyurethane shall
be injected again for sealing.
When the tunnel gate ring is out of the tailskin, the
welding work between ring steel plate, seal steel plate
and embedded steel plates shall be done immediately
to fill the gap between tunnel gate ring and tunnel.
of 3 m higher than the TBM bottom. Steel circumferential plate is arranged along the steel tunnel annulus.
The inner diameter of steel plate is 5 cm larger then
TBM. 18 grout holes are arranged surrounding the
tunnel annulus and inflatable bag is installed in the
tunnel eye.
(2) TBM arriving
When the cutting surface of TBM is close to the concrete wall of tunnel eye, advancing is stopped. Then
pump water in the receiving shaft to the underground
water level. Meanwhile, inject double grout into the
annulus 30 m back from tailskin through the preset
grout hole on the segment to stabilize the asbuilt tunnel and block the water/soil seepage passage between
untreated ground and TBM.
After above work, the TBM starts excavation of
C30 glass fibre reinforced concrete and accesses the
working shaft. The cutting surface accesses into the
working shaft and the cutter head will cut the MU5
cement mortar directly and sit on the mortar layer. During accessing into the working shaft, polyurethane is
injected through the chemical grouting holes.
6.2.2 TBM advancing management

6.2.2.1 Main construction parameters
During TBM construction, the construction parameters shall be defined and adjusted based on theoretical
calculation and actual construction conditions and
monitored data to realize dynamical parameter control
management.
The advancing speed at beginning and before stop
shall not be too fast. The advancing speed shall be
increased gradually to prevent too large starting speed.
During each ring advancing, the advancing speed shall
be as stable as possible to ensure the stability of cutting surface water pressure and smoothness of slurry
supply and discharge pipe. The advancing speed must
be dynamically matching with the annulus grout to
fill the build gap timely. Under normal boring condition, the advancing speed is set as 24 cm/min. If
obstacles varying geological conditions are experienced at the front face, the advancing speed shall be
reduced approximately according to actual conditions.
Based on the theoretical excavation amount calculated from formula and compared to actual excavated
amount which is calculated according to the soil density, slurry discharge flow, slurry discharge density,
slurry supply density and flow, and excavation time, if
the excavation amount is observed too large, the slurry
(3) Tunnel eye sealing and water pumping

When 2/3 of TBM accesses the receiving shaft, water
pumping is started. After pumping the water in the
working shaft, continue the TBM advancing and inject
the grout timely. When the TBM is in the working
shaft, fill air in the inflatable bag in time to make the
inflated bag seal the circumferential gap. Meanwhile,
grouting is performed through the 18 holes on the tunnel annulus. Grout material is polyurethane. After the
gap is fully filled with the grout, the air in inflated bag
39
6.2.3 Quality assurance technical measures for

large tunnel
6.2.3.1 Segment prefabrication
Nine sets steel formwork with high preciseness are
used for segment prefabrication to fulfill the technical requirement to segment such as allowable width
tolerance 0.40 mm, thickness tolerance +3/1 mm,
arc length 1.0 mm, circular surface and end surface
plainness 0.5 mm. In order to control prefabrication preciseness strictly and ensure the production
quality, special laser survey system is introduced
to conduct accurate measurement of segment profile dimension beside traditional survey measurement
tools and segment trial assembly.
Fly ash and slag are mixed in the concrete
for segment prefabrication. Strictly concrete casting,
vibrating and curing procedures are used to control
cracks and achieve the water proofing and durability
requirement.
density, viscosity and cutting face water pressure shall

be checked to ensure the front surface stability.
In order to control the excavated soil amount, the
flow meter and density meter on the slurry circuit shall
be checked periodically. The slurry control parameters
are: density = 1.151.2 g/cm3 , viscosity = 1825 s,
bleeding ratio <5%.
Single type grout is used to inject at 6 locations,
which is controlled by both pressure and grout amount.
The grout pressure is defined as 0.450.6 Mpa. Actual
grout amount is around 110% of theoretical build gap.
20 h-shear strength of grout shall not be less than
800 Pa and 28 day strength shall be above the original
soil strength.
6.2.2.2 Shallow cover construction
At the launching section, the minimum cover depth
is 6.898 m, i.e. 0.447 D, which is extremely shallow.
To ensure the smooth advancing, 12 m soil is placed
above the top. Meanwhile, in order to prevent slurry
blow-out, leakage-blocking agent is mixed in the slurry
and surface condition is closely monitored.
6.2.3.2 Segment assembly

The segment assembly shall satisfy the fitted tunnel
design axis requirement by segment selection (rotation
angle) and meanwhile make the longitudinal joint not
on the same line. During the whole assembly process,
for straight line, the principle is to erect on left and right
at intervals. For curved section, the suitable segment
rotation angle shall be selected based on TBM attitude,
and segment lipping data.
Secondly, the relative dimension between segment
and shield shall be checked to correct the positioning
of each ring segment.
Then, each segment building shall be closely contacted. The ring plane and T joint shall be even.
Finally, strictly control the lipping of ring. When
the segment lipping exceeds the control value, the
rotational angle of segment shall be adjusted timely
to ensure the verticality between segment and tunnel
axis.
6.2.2.3 Crossing the bank of Yangtze River

Before the TBM crossing, the terrain and land feature in the construction surrounding area are collected,
measured and photographed for filing. 155 monitoring points are arranged along the bank in 7 monitoring
sections. During TBM crossing, the pressure is set
according to the water pressure at excavation surface
calculated for each ring. The slurry parameter is also
adjusted timely based on the surface monitoring information. Grease injection at tail skin is performed well
to avoid leakage and synchronous grout amount and
quality are strictly controlled.
6.2.2.4 Adverse geological condition
(1) Shallow gas
When the TBM is crossing the deposit on Pudong
side, methane gas may be experienced in the shallow area. At this time, the ventilation in the tunnel
shall be increased to ensure good ventilation conditions of TBM. The concentration test of methane and
combustible gas are carried out.
6.2.3.3 Floating-resistance of tunnel

Due to the tunnel diameter up to 15 m, the floating
resistance and deformation control during construction for large diameter tunnel are very challenging.
The technical measure is mainly to improve the synchronous grouting management. Mortar type grouting material with cementation property is injected at
multi-points. Furthermore, grout package with certain
strength shall be formed surrounding the tunnel timely
to resist the tunnel upfloating. Meanwhile, the tunnel
axis shall be strictly controlled during construction and
the tight connect between segments shall be improved
to achieve the tunnel-floating resistance.
(2) Lens
Prior to the construction, geological investigation is
carried out to learn the general location of prism. During construction, the TBM is set with suitable speed to
cross the stratum as fast as safely possible.
(3) Bored hole
Due to the tunnel alignment adjustment, 9 geological bored holes will be experienced along the TBM
advancing. During crossing, slurry with large density
is used and polyurethane is injected surrounding the
tunnel after crossing.
6.2.3.4 Ground deformation control

The ground settlement during TBM construction is
mainly contributed by the front surface slurry pressure setup, annulus grouting and shield body tamper.
40
Figure 13. STP system flow chart.
supply pipe has a diameter of 600 mm and discharge

pipe 500 mm. To ensure the long distance slurry supply
velocity of 2.5 m/s and discharge velocity of 4.2 m/s
to avoid slurry settlement in pipe and maintain not too
high pressure in the pipe, one relay pump is arranged
every 1 km. The pressure at pump outlet is controlled
within 10 bar.
Therefore, the ground settlement variation can directly

reflect the TBM construction parameters setup. The
crew can correct the construction parameter based on
settlement monitoring to increase the deformation.
6.2.4 Back-up technology for long distance TBM
construction
6.2.4.1 Slurry treatment and transport
The slurry separation system consists of subsystems
of treatment, conditioning, new slurry generation,
slurry discharge and water supply; with a capacity
of 3,000 m3 /h to fulfill the advancing requirement
of 45 mm/min, as shown in Figure 13. Based on
the geological conditions along the tunnel alignment,
the treatment system selects 2 level treatment methods. The initial treatment uses two rolling shieve to
separate soil with a size of larger than 7 mm. For secondary treatment, firstly grain with a size of large
than 75 is separated by 4 750 mm cyclones and
then grain with a size of large than 40 m is separated
by 12 300 mm cyclones. The slurry spilled at the
top of cyclone is transported to conditioning tank for
reuse. After adjustment, the density of supplied slurry
is 1.051.35 g/cm3 . The maintained optimal value is
between 1.20 and 1.30 and d50 is between 40 and
50 m. The STP system circulation efficiency is up
to 70%. Discharged slurry and waste is transported to
the bardge at riverside by pipes and trucks. The slurry
6.2.4.2 Axis control and construction survey

guidance
Static measurement with GPS control net is used for
surface control survey. For elevation control, GPS elevation fit method is used for elevation transfer. Part of
basic traverse mark every 500 m is selected as main
traverse. In the tunnel, level II subtraverse is used for
the plane control, i.e. construction traverse and control parallel traverse. The control mark has a spacing
of 600900 m. The elevation control survey in tunnel
uses level II. The fixed level mark is arranged with a
spacing of 80 m.
6.2.4.3 Construction ventilation and fire protection
Due to the large diameter, long distance and W longitudinal slope, especially when the TBM is advancing
with a upgrading slope, the heat and humidity generated at the working face can not be discharged naturally
thus concentrate at the working face in a shape of
fog. Meanwhile, heavy trucks for construction material transport also cause a large amount of waste air
41
The wear condition of cutting tools can be indicated

by checking the closed/open status of loops.
The worker accesses the cutting wheel arms from
the center of the main drive. The worker installs the
lowering/ lifting frame (with bolts) and screws it to
the fixing plate of the tool. The fixing plate is then
unscrewed. The worker will then lower the tool using
the frame (with bolts). The pressure-tight gate will
be closed down. The worn out tool shall be then
exchanged with a new one. The tool will be lifted to
position behind the gate. The gate will be opened. Then
the tool will be put in its final position. The fixing
plate is then screwed to the tool support. The frame is
transferred to the next tool.
in the tunnel. Bad environment will have unfavorable

influence on TBM equipment and crew, and also affect
the smooth progressing of survey activity.
During construction stage, 2 special axial fans
(SDF-No18) are arranged on the surface to provide
fresh air to the space below road deck in the tunnel,
then the relay fan and ventilation system equipped
on the gantry will transport the fresh air to working surface. Meanwhile, other ventilation equipment
on the gantry provides fresh air to main secondary
equipments of TBM such as transformer, hydraulic
equipment and electrical installations.
Adequate fire extinguishers are arranged in the
shield and gantry and also oxygen, poisonous gas
protection mask are equipped. Fire extinguisher is
equipped on each transport truck. Safety staff is
equipped with portable gas analysis device for check
the air quality in tunnel every day.
6.2.5.3 Tail seal and steel wire brush replacement

When the leakage is experienced at tail skin, and
steel brush is defined to be replaced necessarily, open
the emergency sealing and erect special segments.
Strengthen the surrounding soil at tail skin with freezing method and then replace first 2 or 3 rows steel
brush.
6.2.4.4 Material transport

Segment, grout and prefabricated elements, etc are
transported to the working area by special trucks from
ramp area, through cut & cover tunnel and road deck
which is constructed synchronously. Truck transport
can avoid the derailing problems during traditional
electrical truck transport. Furthermore, the truck has
two locos, so the transport efficiency is high.
Prefabricated road element is transported to the
gantry 2 by trucks and then lifted and erected by the
crane on the bridge beam. Segments are transported to
gantry 2 and then transferred to the segment feeder by
the crane on the bridge beam and then transported to
erection area.
6.3
Synchronous construction of road deck
The synchronous construction of road structure

includes erection of road element, segment roughening and drilling for inserting rebar, prefabrication of
two side ballast, insitu cast corbel and road deck on
two sides. According the variation and trend of asbuilt
ring deformation and settlement, and the construction
progress of 12 m (6 rings) per day and based on the
requirement of deformation joint arrangement every
30 m, the construction is organized and arranged as
flowing operation every 15 m. As shown in Figure 14,
the basic construction procedure is as follows:
6.2.5
Critical equipment examination and

replacement technology
6.2.5.1 Main bearing sealing
Four supersonic sensors are installed in the seal
arrangement for monitoring the main seal wear condition. Once the abrasion reaches certain value or grease
leakage is monitored in the tank, it indicates the main
seal needs to be rotated to another oritentation.
Once the seal wear is observed beyond preset value,
the surface could be moved to ensure the replacement
of main bearing seal. During replacing, the slurry in
the chamber must be drained and provide effective support to excavation face. The operation staff shall go
to the slurry camber to replace the seal under certain
pressure.
Road element installation, 25 rings later than segment erection.

Segment roughening includes the junction surface
between ballast and segment and segment inner
surface at corbel. The insert bar placing includes
the
16 bar at ballast and
20 bar at corbel. The
roughening works at ballast position is carried out at
gantry 2, and the roughening operation platform at
corbel is fixed to gantry 2. Insert bar placing is following gantry 3. The roughing machine is equipped
with dust suction facility which can eliminate the
dust to maximum extent.
Reinforcement placing, formwork erection and
concrete casting for ballast is carried out at 15 m
behind the gantry 3 and 15 m more behind for corbel, and then another 15 m for road deck. Road deck
concrete casting works are located at 250 m300 m
from the segment erection area. After casting, the
curing with frame lasts 3 days and formwork is
removed on the 4th day. After 28 days curing, the
road deck can be open to traffic. During curing, the
6.2.5.2 Abrasion measurement and replacement of

cutting tools
The system will be installed on 8 selected scraper positions as well as on two bucket positions. It will be
connected to a plug at the rear of the cutting wheel
to allow for simple condition diagnosis from a readout device. Conductor loop is embedded in the device.
42
Figure 14. Synchronous construction flow chart.
the frozen soil is thawed by section. Based on the

informational monitoring system, the soil temperature
and settlement variation is monitored. Grouting pipe is
arranged at shallow and deeper area for dense grouting. The overall principle for thawing is to thaw the
bottom part, then middle part, and lastly the top part,
as shown in Figure 15. When thawing by section is
done in sequence, one section is being thawed and subsequent sections maintain the freezing for the purpose
of maintaining the cross passage structure and main
tunnel as an integrated part thus settlement avoidance
before the section grouted.
road deck area is separated. Concrete mixing truck

is used for concrete casting.
6.4 Cross passage construction

The cross passage which connects the two main tunnels has a length of around 15 m and diameter of 5 m.
The construction will be by freezing method for soil
strengthening and mining method for excavation.
The freezing holes are arranged as inside and outside rows which are drilled from two sides. The
freezing is done from one side or both sides. Inside row
holes are drilled from upchainage tunnel, 22 in total
and outside row holes are drilled from downchainge
tunnel, 18 in total.
Mining method will be used for excavation by area
division. Firstly, pilot with a horn opening is excavated, and then the cross passage is excavated to design
dimension. The fullsection excavation is done with a
step of 0.6 m or 0.8 m.
When the main structure concrete strength reaches
75%, enforced thawing will be carried out. The hot
brine for thawing circulates in the freezing pipe and
6.5 Land connections construction

The profile dimension of working shaft is 22.4 49 m,
with a depth of 25 m. 1.0 m thick diaphragm with a
depth of 45 m is used for retaining structure. Open cut
is used for excavation. The support system consists of
5 layers reinforced concrete and 1 layer steel support.
Inside the pit, 3 m below the bottom, injection is done
interval to make the strength not lower than 1.2 MPa.
13.516.0 m outside the working shaft is treated. For
43
CONCLUSION
During the process from planning to implementation, Shanghai Yangtze River Tunnel has experienced
various challenges. Technical support of tunnel construction from China and abroad is provided. With
indenpendently developed and owned IPR and featured TBM tunnel construction theory and core technology is established, forming the core technology of
large and long river-crossing TBM tunnel in China.
Special technical issues such as lining structure design
of extremely large tunnel, long distance TBM construction and hazard prevention system for long and
large tunnel achieve to be internally state-of-art. Relevant standards, codes, guidance, specification and
patent technology are developed to improve the technical system of tunnel construction in China and upgrade
the internal competence of tunnel engineering.
REFERENCES
Cao, W.X. et al. 2006. ShanghaiYangtze River Tunnel Project
design. Shanghai Construction Science and Technology 5:
26.
Chen, X.K. & Huang, Z.H. 2007. Shanghai Yangtze River
Tunnel TBM cutting tools wear detection and replacement
technology. The 3rd Shanghai International Tunneling
Symposium Proceedings: Underground project construction and risk provision technology: 152157. Tongji
University Publication Company.
He, R. & Wang, J.Y. Shanghai Yangtze River Tunnel synchronous construction method statement. The 3rd Shanghai International Tunneling Symposium Proceedings:
Underground project construction and risk provision technology: 168177.Tongji University Publication Company.
Sun, J. & Chen, X.K. 2007. Discussion of TBM selection
for Shanghai Yangtze River Tunnel. The 3rd Shanghai
International Tunneling Symposium Proceedings: Underground project construction and risk provision technology: 9198. Tongji University Publication Company.
Yu, Y.M. & Tang, Z.H. 2007. Shanghai Yangtze River Tunnel construction survey technology. The 3rd Shanghai
Zhang, J.J. et al. 2007. Shanghai Yangtze River Tunnel TBM
launching construction technology. The 3rd Shanghai
Figure 15. Divided thawing area of cross passage.
diaphragm at the TBM accessing into the receiving

shaft, GFPR is used instead of normal reinforcement
so that the TBM can cut the retaining wall directly
and thus avoid the reinforcement cutting and tunnel
eye concrete removal, which simplifies the construction procedure, accelerates construction progress and
reduces the construction risk.
The excavation depth of pit for Pudong cut-andcover is 23.19.9 m, and Changxing cut-and-cover
17.2 m8.4 m. According to the excavation depth,
diaphragm with thickness of 1.0 m, 0.8 m and 0.6 m is
selected respectively. The support system is composed
of reinforced concrete support and steel support. 3 m
underneath the pit bottom is strengthened by rotating injection and also the junction between working
shaft and cut-and-cut outside the pit to ensure the pit
excavation stability.
The ramp is open cut with a slope of 1:3. The slope
is protected through green planting in the reinforced
concrete grid which is anchored in soil by anchors to
prevent from sliding. In order to avoid slope sliding,
the slope is strengthened by cement mixed piles with
a diameter of 700 mm.
44
Underground construction in decomposed residual soils

I.M. Lee
Department of Civil Engineering, Korea University, Seoul, Korea
G.C. Cho
Department of Civil and Environmental Engineering, KAIST, Daejeon, Korea
ABSTRACT: Large scale underground construction projects, including subway construction projects in six
major cities, have been ongoing in Korea, where residual and granite soils are the most common soil type.
Characteristics of decomposed granite soils are different from those of pure sand and/or clay. This paper presents
an overview of geotechnical aspects of underground construction in urban areas where mostly decomposed
residual soils are present, focusing on mechanical properties, apparent earth pressure, effect of groundwater,
and effect of spatial variability in geotechnical properties. Although several important aspects are theoretically, numerically, and experimentally discussed herein, it remains a challenge to fully understand residual
soils, particularly in relation to the practice of underground construction, because of their complexity and
richness.
INTRODUCTION
earth pressure, effect of ground water, and effect of

spatial variability in geotechnical properties.
Large scale subway construction projects have been

ongoing in six major cities in Korean peninsular.
In particular, underground construction work of subway line No.9 is being carried out under the Seoul
Metropolitan Government along with extension of
subway lines No.7 and No.3. Construction of a new
subway line (Line No.2) will be launched in Incheon
this year and will be finished before the Asian Games
are held in 2014.
Residual and granite soils are the most common soil
type in Korea. Characteristics of decomposed granite
soils are different from those of sand and/or clay. Their
mechanical properties and behaviors vary depending
on the parent rock types and weathering processes.
Moreover, the profile of the ground in Korea is generally not uniform, isotropic, or homogeneous; multilayered conditions are common, with ground being
composed of successive layers of fill and/or sedimentary layers, weathered residual soils, and soft to hard
rock. Therefore, conventional/classic soil mechanics
cannot be directly applied to these ground conditions.
In order to provide data and methodologies to enhance
underground construction in areas characterized by the
predominant presence of decomposed residual soils,
this paper presents an overview of geotechnical aspects
of underground construction in urban areas where
decomposed residual soils are the main ground component, focusing on mechanical properties, apparent
2
2.1
MECHANICAL PROPERTIES OF RESIDUAL

SOILS
General
We have intensively studied the characteristics of the

following two residual soils: Shinnae-dong and Poidong soils. The characteristics of the two soils are
documented in Table 1 (i.e., the type of mineral, compaction properties, plasticity, and soil classification).
The parent rock of the Shinnae-dong soil is a granite while that of the Poi-dong soil is a banded biotite
gneiss. The Shinae-dong soil is closer to a cohesionless
soil, predominantly consisting of primary minerals,
with only about 10% of fine particles. On the other
hand, the Poi-dong soil shows clay-like characteristics
due to a large percentage of fine particles and secondary minerals. The particle size distributions of the
two soils are shown in Figure 1.
Lee (1991) studied the behavior of a Bulam soil
in his Ph.D. dissertation. The characteristics of the
Bulam soil are similar to those of the Shinna-dong
soil, since two areas are very close to each other and
share the same rock origin. Kim (1993) studied the
mechanical behavior of Andong and Kimchun soils.
These soils are classified as SM in the Unified Soil
45
Table 1.
Characteristics of two residual soils.
Characteristics
Primary minerals
Quartz (%)
Feldspar (%)
Mica (%)
Secondary minerals
Kaolinite (%)
Illite (%)
Vermiculite (%)
Chlorite (%)
Montmorilonite (%)
Porosity
Maximum dry density (kN/m3 )
Percent passing #200 sieve (%)
Plasticity
LI
PL
PI
Specific gravity
USCS
Poi-dong
Shinnae-dong
17.7
15.0
9.8
33.3
50.0
9.0
23.5
20.7
8.4
4.5
6.0
0.409
16.68
47.36
34.0
19.84
14.16
2.74
SC
2.0
0.358
18.64
10.05
Nonplastic
2.65
SW-SM
Figure 2. Peak internal friction angle versus fine contents

passing the #200 sieve.
Note: LI = liquidity index; PL = plastic limit; PI = plasticity

index; USCS = Unified Soil Classification System.
Figure 1. Particle size distribution of two residual soils.
Classification System with the percent passing a #200

sieve being 1417%, and are also included in the
following discussion on mechanical characteristics of
residual soils.
2.2
Figure 3. Failure envelopes at different matric suctions.
Lee et al. (2002) performed shear tests on unsaturated residual soils and found that the apparent
cohesion can increase from 20 kPa at a saturated condition (i.e., matric suction = 0 kPa) up to 200 kPa at an
unsaturated condition (i.e., matric suction = 400 kPa).
As an unsaturated soil is re-saturated, its apparent cohesion can be eliminated. Thus, during tunnel
construction, cohesion loss can be induced by resaturation (e.g., seepage hindrance, drainage clogging
and groundwater change), and may result in tunnel
face instability. Utilizing the limit equilibrium analysis proposed by Leca & Dormieux (1990), as shown in
Figure 4, the required support pressure to stabilize the
Strength characteristics
Figure 2 presents a summary of the peak internal friction angle of each residual soil. The internal friction
angle decreases with an increase of fine contents.
Because of capillarity, partial saturation affects the
strength of residual soils. Lee et al. (2005) performed
triaxial tests to obtain the strength properties of unsaturated residual soils. Figure 3 presents typical results
of failure envelopes at different matric suctions. The
internal friction angle as well as the apparent cohesion
increase with an increase in the matric suction (i.e.,
when the soil is unsaturated).
46
Table 2. Typical material properties.
Layer
E
(MPa)
(kN/m3 )
K0
c
(kPa)
( )
Filling
Residual soil
Weather rock
Soft rock
19.6
29.4
196
981
0.35
0.33
0.23
0.2
18.6
18.6
21.6
23.5
0.5
0.5
0.5
0.7
0
49.1
98.1
196
35
35
35
40
Notation: E =Youngs modulus, = Poissons ratio, = unit

weight of soil, K0 = lateral earth pressure coefficient,
c = apparent cohesion, and = internal friction angle.
Figure 4. Collapse mechanism of a tunnel face with two

conical blocks.
Figure 5. Ground condition for limit equilibrium analysis.
tunnel face can be calculated by replacing the specific

apparent cohesion with a value of zero. As an example,
the ground condition of a site is shown in Figure 5. The
required support pressure increases from zero up to
= 9.0 kPa as the cohesion decreases from 120 kPa
to 0 kPa. The results show that the apparent cohesion
is a key factor in tunnel face stability.
2.3
Figure 6. Ground conditions of the Busan subway site.
Deformation characteristics
One of the most difficult tasks in geotechnical engineering is estimating the deformation-related soil
properties properly. Typical material properties commonly used at the design stage in Korea are summarized in Table 2.
Cho et al. (2006) proposed an analytical method
to estimate soil parameters from relative convergence
measurements. As an example, the ground conditions
of the Busan subway site are shown in Figure 6. Initial estimates are taken from Table 2. The crown and
sidewall convergence data measured by a tape extensometer, presented in Figure 7, are used as observed
values. Results obtained from back-analyses are summarized in Table 3 along with initial estimates. There
is a large discrepancy between the initial input properties and the properties obtained from back-analyses.
In particular, the initial inputs of Youngs modulus
and earth pressure coefficient at rest of the residual
Figure 7. Relative convergence behind tunnel face.
Table 3.
Results of parameter estimation.
Properties
Er (MPa)
Ew (MPa)
Kow
Initial input
Back-analysis
29.4
83.3
196
210
0.5
0.76
Notation: Er and Ew =Youngs moduli of the residual soil and

weathered rock, respectively; and Kow = the earth pressure
coefficient of the weathered rock.
47
Figure 9. Maximum apparent earth pressure versus Ka H.
3.2 Apparent earth pressure in a vertical shaft

Figure 8. Apparent earth pressure distribution for braced
and anchored walls.
It is well known that the earth pressure acting on a vertical shaft is less than that on a retaining wall, because
of the three dimensional arching effect. The existing
equations of earth pressures acting on vertical shafts
consider only either purely cohesionless or cohesive
soils. These solutions are not directly applicable to
estimation of earth pressures for multi-layered ground.
Lee et al. (2007) proposed an equation to estimate
earth pressures in multi-layered ground, assuming
that the failure shape is conical, as shown in Figure 10(a). For equilibrium of horizontal forces and
vertical forces, as shown in Figure 10(b), the earth
pressure (Pi ) can be expressed as follows
soil are too small. This example clearly demonstrates

the importance of the observational method in tunnel
engineering.
3 APPARENT EARTH PRESSURE

3.1 Apparent earth pressure in braced and
anchored walls
Design of an in situ wall system requires a lateral earth
pressure distribution behind the wall to estimate support loads and wall bending moments. One of the most
well-known apparent earth pressures is that proposed
by Peck (1969); however, his suggestion is only applicable to either ground that is purely sand and/or clay,
and cannot be directly applied to cohesive soils that
have cohesion as well as an internal friction angle, or
to multilayered ground conditions.
Many Korean researchers have attempted to collect
field data to propose the earth pressure in multilayered
ground (for example, Lee & Jeon 1993, Yoo 2001).
Figure 8 presents a typical set of results, showing the
apparent earth pressure distribution of a 33m deep
excavation site along with the distribution proposed
by Peck (1969) (Yoo 2001). The actual (measured)
earth pressure is about 68 to 83% of Peaks. Figure 9
shows the maximum earth pressures obtained from
62 excavation sites. In this data, the weighted average values of the internal friction angle and the unit
weight of soil are used for the multilayer ground. The
average value of the measured apparent pressures is
approximately 75% of Pecks suggestion (i.e., earth
pressure = 0.65 Ka H).
and
where Kw = the coefficient of radial earth pressure, = the coefficient of tangential earth pressure,
and = the wall friction angle. Figure 11 presents
schematic drawings of a construction site in multilayered ground and three vertical shafts along with the
locations of measuring instruments. Earth pressures
are measured at different shafts. The earth pressures
calculated from Eq. (1) are compared with the measured values, as shown in Figure 12. The measured
earth pressures are even smaller than those obtained
from the theoretical equation.
48
Figure 11. Schematic drawings of a construction site in

multi layered ground Three shafts with locations of measuring instruments.
on such collapse mechanisms reveals the following

common features:
1. Thin soil/rock cover and/or mixed faced ground
conditions including decomposed granite soils;
2. Collapse initiated at the tunnel shoulder during
excavation of the upper half of the tunnel section;
and
3. A considerable amount of groundwater inflow with
soil.
Figure 10. Derivation of earth pressure in vertical shaft.
4
4.1
EFFECT OF GROUNDWATER
In particular, it is observed in these sites that tunnel face collapses always occurred along with seepage
ahead of the tunnel face.
Lee et al. (2003) modified the upper bound solution originally proposed by Leca & Dormieux (1990),
taking into account seepage forces in a stability assessment of a tunnel face (refer to Figure 4). The horizontal
components of seepage pressures acting on the tunnel
face, S.F , can be simply considered as an external load
Effect of seepage pressure
Unexpected groundwater inflow and seepage forces

often cause tunnel failures during construction. Shin
et al. (2006) presented and reviewed five cave-in collapses that occurred while constructing Line No.5 of
the Seoul Metropolitan subway. Figure 13 shows the
general features of the collapses and failure details
are summarized in Table 4. A comprehensive review
49
Figure 13. Cave-in collapses in the Seoul Subway Line

No.5.
in the opposite direction of the support pressure T .

A modified upper bound solution with consideration
of seepage forces becomes
where P is the surcharge, c is the unconfined compressive strength of the soil, T is the required support
pressure applied to the tunnel face, S.F is the seepage pressure acting on the tunnel face, Kp is Rankines
earth pressure coefficient for passive failure, is the
unit weight of soil, D is the tunnel diameter, and Ns
and N are the weighting coefficients, respectively.
When tunnel excavation is performed below the
groundwater level, the stress condition in front of the
tunnel face becomes the summation of the effective
stress and the seepage pressure. The effective stress
can be calculated by the upper bound solution while the
seepage pressure can be obtained from numerical analyses. The effective support pressure at the tunnel face
can be obtained by Eq. (4) with use of the submerged
unit weight sub instead of .
As an example analysis, consider a virtual tunnel with a diameter D driven horizontally under a
depth C, as shown in Figure 14. Ground material
properties used for the analysis are c = 0, = 35 ,
and sub = 5.4 KN/m3 . Figure 15 shows the total head
distribution around the tunnel face, determined by
seepage analyses, and the failure zone, estimated from
a limit equilibrium analysis. The seepage pressure is
calculated by using J = iw A, where i is the hydraulic
gradient and A is the area. The total support pressure
is then obtained by summing up the effective support
pressure and the seepage pressure. Figure 16 shows the
support pressure change with variation of the H/D ratio
(For the case of a dry condition, the dry unit weight
d = 15.2 KN/m3 is used for the analysis). The results
suggest that the total support pressure is little affected
by the tunnel depth and increases significantly with
Figure 12. Earth pressures measured at vertical shafts.
50
Table 4.
Collapse mechanisms, damage, and remedial work.
Case
Failure mechanism
17 Nov 1991, 18:50 : blasting

21:05: total collapse (1,000 m3 )
thin weathered rock cover
inflow of soil and groundwater
weathered granite (WG) at the face
very close to an existing stream
27 Nov 1991, 10:40: blasting
16:00 :rock falls at the face
22:00: soil and water inflow (D = 25 m)
28 Nov 1991,
03:20 :additional collapse (D = 20 m)
WG at the face
permeability: (1.0 104
2.0 105 cm/sec)
11 Feb 1992, road header excavation
04:30: raveling
significant inflow of ground
water (100130 /min
about 4.5 ton of soil flew into
face: D = 10 m
heavily WG at the face
7 Jan 1993, 03:30: blasting
collapse after removing materials
collapse size: 0.7 m 1.2 m
started at the left side of crown
soil inflow: 900 m3 , groundwater:
300 /min
WG & DGS at the face
1 Feb 1993, ring cut
08:30 : rock fall and collapse
(oval shape D = 1030 m)
inflow of soil with groundwater
alluvium & DGS at the face
beneath an existing stream
Damage
Remedial work
no human casualties
2-1ane road collapsed
stop of gas supply
(5,000 households)
damage to lighting poles
and traffic light poles
no human casualties
80 households evacuated
electricity and water
main collapsed
(3-story bldg) slipped
into crater
soil dumping immediately after

collapse
face shotcrete and invert Close
(t = 1.5 2.0 m)
cement milk grouting and curtain
wall grouting
grouting: cement mortal cement
milk chemical grout
lowering of groundwater
level(3 m/day)
pumping/fore poling for
re-excavation
remedial cost: $4.5million
no human casualties
dumping soils into collapsed
4-lane road collapsed
area (240 tons)
service culvert (6 m 3 m) dumping ready-mixedexposed (including 154 kv
concrete (105 tons)
electricity cable)
grouting
passengers delay
no human casualties
grouting: cement mortal, cement
2-lane road collapsed
milk and LW(160 holes)
supply stop of water
mortal injection beneath sewer culvert
main(
= 200 mm)
reduce inflow of groundwater
sewer culverts were broken
using chemical grouting
40 households evacuated
no human casualties
6 items of excavation
equipment were buried
soil dumping into collapsed

area (5,500 m3 )
chemical grouting (76 holes)
compaction grouting (300 holes)
remedial cost: $1.7 million
Figure 15. Hydraulic head distribution and failure zone.

Figure 14. Dimensional condition for seepage analysis.
effect of seepage plays an important role in tunnel face

stability problems.
an increase in the groundwater level ratio. As the total

support pressure is related to the tunnel face stability, the seepage force seriously affects the tunnel face
stability. While the effective overburden pressure is
reduced slightly by the arching effect during tunnel
excavation, the seepage pressure remains at the same
level during tunnel excavation. This explains why the
4.2 Particle transport characteristics of granite

residual soils
A soil is said to be internally stable if it is self-filtering
and if its fine particles do not move/migrate through
51
Figure 16. Change of support pressure with variation of the

H/D Ratio.
Figure 18. Cumulative mass versus (a) time and (b) pore
volume for Shinnae-dong soil.
the cell is characterized with respect to its turbidity

(in terms of nephelometric turbidity units, NTUs) and
particle size distribution.
The cumulative mass of particles in the effluent
eroded from the base soils is plotted with respect to
time in Figures 18(a) & 19(a) for the two residual soil
types. Some important differences can be observed in
the internal erosion behavior of the two soil types. The
Shinnae-dong soil exhibits almost the same rate of erosion during the initial stages of the experiment for the
three different flow rates used (Figure 18a). Particle
redeposition in the soil sample appears to compensate
for the increased erodibility at higher flow rates. This is
even more apparent when the cumulative mass is plotted in terms of pore volume (Figure 18b). It is seen that
at low flow rates the erosion rates are higher, because
of the reduced particle deposition. There appears to be
a maximum limit for the cumulative mass of internal
erosion for each flow rate, beyond which the soil protected itself from further erosion, perhaps through the
formation of a filter bridge. For the relatively cohesive
Poi-dong soil (Figure 19a & 19b), self-protection
due to particle redeposition is not apparent. There is
no cap on the maximum eroded quantities during the
period of testing.
In contrast to the internal erosion behavior, the surface erosion from the two samples (as observed in
experiments where erosion is induced in a cylindrical hole) follows almost linear trends, with the rates
Figure 17. Schematic drawings of experimental set-up.
the pores of its own coarser fraction. Previous investigations into the internal stability of cohesionless
soils suggest that soils with a uniformity coefficient
(Cu ) > 20 and with concave upward grain size distributions tend to be internally unstable (Lee et al. 2002).
Most residual soils in Korea, including those listed
in Table 1, have uniformity coefficients much greater
than 20, suggesting that they are internally unstable.
Lee et al. (2002) studied the nature of particle
transport and erosion in residual soils. Two types
of residual soils introduced in Section 2 are used:
Shinnae-dong soil and Poi-dong soil.The experimental
setup is shown in Figure 17. In selected experiments,
a cylindrical hole 7 mm in diameter is drilled into the
compacted specimens to induce erosion only in the
hole and simulate surface erosion of the soils. An electronic pump is used to achieve a constant flow rate of
the influent from the water tank. The effluent from
52
Figure 20. Grain-size distribution of soils and grouts.
Table 5.
tests.
Soil and grout properties used in chamber injection
Material
Soil A
Soil B
Fine cement
Quick setting agent
Fine cement + Quick
setting agent
D10
(mm)
D15
(mm)
d85
(m)
d95
(m)
0.60
2.10
0.64
2.22
16
37
20
27
70
39
32
111
Figure 19. Cumulative mass versus (a) time, and (b) pore
volumes, for Poi-dong soils.
case of N > 25, the following requirement should be

additionally satisfied for the soil to be groutable:
of erosion increasing as the flow rate increases. The
rates of erosion are also considerably higher than those
obtained in the internal erosion experiments discussed
above.
Particle transport characteristics of residual soils
might be among the factors that result in the instability
of underground structures.
4.3
where D10 is the particle size of base soils corresponding to 10% finer and d95 is the particle size of
grouts corresponding to 95% finer. Kim et al. (2007)
performed pilot-scale chamber injection tests to investigate the groutability of two granular soils that satisfy
the groutability criteria proposed by Burwell.
The grain-size distributions of soils and grouts are
shown in Figure 20 and their properties are summarized inTable 5.The experimental set-up for pilot-scale
chamber injection tests is shown in Figure 21. Typical results of the experiments are shown in Figure 22.
Although the N value of the soil A (N = 32) is greater
than 25, the grout could not be sufficiently injected into
soil A. Meanwhile, groutability is fairly good for soil B
(N = 111). These results suggest that the consideration
of filtration phenomena is indispensable to reasonably evaluating the potential of grout penetration. The
N value of the Shinnae-dong soil shown in Figure 1 is
2.6 and that of the Poi-dong soil is 0.2. These values
mean that penetration grouting in these soils is almost
impossible. Therefore, finding an appropriate grouting method has presented a considerable challenge in
granite residual soils.
Difficulties in penetration grouting
Tunnelling works in soft ground frequently require

grouting technology, either to prevent groundwater
or to improve mechanical properties of the ground.
However, grouting is not available in many cases in
decomposed residual soils due to low groutability. Burwell defines the groutability (N ) of suspension grouts
by the following simple equation (Kim et al. 2007):
where D15 is the particle size of base soils corresponding to 15% finer and d85 is the particle size
of grouts corresponding to 85% finer. If N is larger
than 25, grout can be successfully injected into the
soil formation. However, Burwell notes that even in
53
Figure 23. Hydraulic head at soil-grouting interface

depending on the permeability ratio.
is applied around the tunnel, a loss of hydraulic

heads occurs in the grouting zone around the tunnel;
this reduces the seepage force acting on the shotcrete
lining, and results in a favorable ground reaction.
Following Darcys continuity equation, the hydraulic
head acting on the soil-grouting interface can be
written as
Figure 21. Experimental set-up for pilot-scale chamber

injection test.
where HI is the total head at the soil-grouting interface,

HT is the total head of a site, Lg is the thickness of the
grouting, Ls is the length across which water travels
through the soil media, and is the permeability ratio
between the soil and grouting area (i.e., = Kg /Ks ).
Figure 23 shows the variation of the hydraulic head at
the soil-grouting interface with different permeability ratios. As the permeability ratio decreases and the
grouting thickness increases, the hydraulic head acting
on the interface increases.
Finite element analyses were performed in order to
explore the effect of grouting on the ground reaction
with consideration of seepage. Seepage force acting
on the grouting-soil interface can be modeled by fully
coupled mechanical-hydraulic analyses, as shown in
Figure 24. Material properties used in this analysis are
summarized in Table 6. It was assumed that shotcrete
is not applied, the groundwater flow is in a steady-state
condition, the grouting thickness is 1 m, and the permeability ratio is = 0.1. Four cases were simulated
numerically: 1) Grouting with seepage; 2) Grouting
without seepage; 3) No grouting with seepage; and
4) No grouting without seepage. Figure 25 presents
the effect of grouting and seepage force on the ground
reaction curve as given by the numerical analysis
results. The case of seepage force without grouting
Figure 22. Maximum injection volume with injection

pressure.
4.4
Ground reaction curve with consideration of

seepage forces and grouting
Theoretical analyses of seepage around tunnels suggest that a loss in hydraulic heads occurs at the
shotcrete lining and concentration of seepage force
at the shotcrete lining in the radial direction induces
unfavorable ground reaction (Shin, 2007). Thus, when
seepage problems are anticipated during tunnel construction, proper grouting around tunnels can provide
effective reduction of seepage force acting on the
shotcrete lining and also increases the stiffness and
strength of the surrounding ground. When grouting
54
though grouting is not applied. This means that the

seepage force significantly affects the ground reaction behavior. In the case where grouting is applied,
unfavorable ground reactions induced by the seepage
force could be considerably reduced.
5
5.1
Material properties used in numerical simulation.
Weathered soil
Grouted zone
E
(MPa)
(kN/m3 )
K0
c
(kPa)
( )
50
500
0.35
0.33
18.64
18.64
0.5
0.5
10
100
35
35
Experimental study on time-dependent

characteristics of grouted residual soil
In Korea, in conventional tunnelling in residual soils,

pre-reinforcement (grout injection) is typically applied
ahead of the tunnel face to enhance the construction safety. In addition, a 1 to 2 day time interval
is given between one face and the next face. During this time interval, it is known that changes in
the material properties occur due to effects of the
curing of the grouting material. However, the stiffness and strength at 28 curing days after the grout
injection are generally applied as the material properties for pre-reinforced zones in the design stage
without considering the effect of the time-dependent
behavior of the injected grout material. Thus, this
paper present a new method to characterize the timedependent behavior of pre-reinforced zones around
a large-section tunnel in residual soil using elastic
waves and to consider time-dependent characteristics
in numerical modeling for tunnel design (Song, 2007).
Figure 26 presents schematic drawings of the experimental setup for investigation of the time-dependent
characteristics of grouted residual soils: (a) Setup
for elastic wave measurements; (b) Setup for shear
strength parameter measurements. Bimorph bender
elements were installed in the testing device and used
to send and receive P- and S-waves (Figure 26a). The
specimens were prepared by mixing a residual soil with
5% cement (by weight; the cement-water ratio is the
same as that used in the field).
Figure 27 shows typical results for the elastic wave
velocity according to the curing time when the normal stress is n = 160 kPa. The results show that the
wave velocity increases drastically according to the
curing time and is almost constant after 7 days. P-wave
velocity is faster than S-wave velocity and Poissons
ratio can be readily determined from the two wave
velocities.
Figure 28 shows the time-dependent characteristics of shear strength parameters obtained from the
direct shear test. As shown in Figure 28(a), the friction angle does not change in accordance with the
curing time. On the other hand, it is apparent that the
cohesion increases with the curing time; after a certain amount of curing time the cohesion converges, as
shown in Figure 28(b). It is deduced that the bonding of cement increases the cohesion and, after with
Figure 24. Finite element model for seepage force analysis

with consideration of grouting.
Table 6.
CHARACTERIZATION AND MODELING OF

GROUTED RESIDUAL SOIL
Figure 25. Effect of grouting and seepage force on ground

reaction curve (i.e., = 0.1).
yields a very unfavorable ground reaction curve, which

induces a large deformation and requires substantial
internal support. However, if the seepage force is not
considered, the ground reacts almost elastically even
55
Friction angle ()
50
40
30
20
10
0
0
3
4
5
Curing time (Days)
(a) Friction angle
Cohesion (kPa)
250
200
150
100
50
0
2
3
4
5
Curing time (Days)
(b) Cohesion
Figure 28. Time-dependent characteristics of shear strength
parameters.
where , , A, and B are the fitting parameters and

t is the curing time. These fitting parameters can
be determined by best-fitting the experimental data
with Eq. (8) and Eq. (9). Also, the shear strength
and strength parameters (i.e., the cohesion and friction angle) can be uniquely correlated to the elastic
wave velocities.
Figure 26. Experimental setup for investigation of

time-dependent characteristics of grouted residual soils.
Velocity (m/sec)
1400
1200
1000
800
5.2 Numerical simulation of time-dependent

characteristics of grouted residual soil
600
400
P-wave
200
S-wave
10
15
20
Curing time (Days)
25
The construction of underground space in residual

soil entails many risk factors such as difficulties in
predicting arching effects and determination of various uncertain underground properties. Researchers
have suggested various techniques for auxiliary support systems such as the reinforced protective umbrella
method (RPUM), which has the advantage of combining a modern forepoling system with a grouting
injection method (Barisone, 1982). This method is
used for pre-reinforcement design before the underground excavation: not only for small section tunneling within weathered and crashed zones, but also for
large underground spaces. In addition, to decrease the
risk of a collapse or failure in large excavation caverns,
researchers have developed various techniques and
construction methods. Some examples include: a tunneling method using an advanced reinforcing system
where a double steel pipe is used for water-proofing
30
Figure 27. Elastic wave velocity according to curing time

(n = 160 kPa).
the elapse of time, the cohesion maintains a uniform

value with the end of cementation. The early stage of
this phenomenon is controlled by the normal stress,
but as curing time increases the cementation controls
the friction angle and cohesion.
The wave velocity and cohesion of grouted residual
soils can be respectively correlated with the curing
time as follows:
56
Normalized Vertical Displacement
1.2
1D Stiffness and Strength
Time-Dependent Stiffness and Strength
1
0.8
0.6
0.4
0.2
0
10
15
20
25
Excavation Length (m)
30
Normalized Horizontal Displacement
(a) Vertical displacement on a tunnel portal
Figure 29. 3D tunnel model and time-dependent material

properties of the pre-reinforced zone after 12m excavation.
1.00
0.95
0.90
Time-Dependent Stiffness and Strength
0.85
0.80
10
15
20
Excavation Length (m)
25
(b) Horizontal displacement on a tunnel face
and a urethane injection is used for reinforcement; the

Trevi jet method, which involves constructing an archshell structure around a tunnel crown with cement
grout; and steel pipe reinforced multi-step grouting,
where a beam arch is constructed around the tunnel
crown with large diameter steel pipes, and multilayer
cement grouting injection is employed.
Three dimensional FE analyses were performed to
examine the time-dependent behavior of the grouted
zone. The results obtained from laboratory tests were
applied to a numerical simulation of a tunnel, taking
into account its construction sequence. Figure 29(a)
shows a simulated 3D four-lane tunnel model, where
the same stress state and stress level as used in the
experiment were assumed. Figure 29(b) shows the
time-dependent elastic modulus and cohesion values
obtained from the experimental study as well as those
used in the numerical analysis.
The time-dependent behavior of a pre-reinforced
zone can be modeled using the following procedure.
The material properties (i.e., stiffness and strength) of
the pre-reinforced zone are considered as the boundary
conditions from Day 1 to Day 28. The registered initial boundary conditions are applied to a pre-assigned
mesh in pre-reinforcement construction. The boundary conditions are then updated according to the field
construction sequence.
For a quantitative analysis, the displacements of
each case are normalized with the results of a pipeonly case. Figure 30(a) shows the normalized vertical
Figure 30. Variation of normalized displacement.
displacement at the portal. The trend of the normalized

vertical displacement curve for the time-dependent
condition is similar to that of the one day curing
case within the initial excavation section (<8 m). As
the excavation continues, the results of the timedependent condition become similar to those of the
2 3 days curing case, until vertical displacement
eventually converges. The stiffness and strength of the
pre-reinforced zone for the 1 2 days curing case are
roughly 30 50% of those of the 28 days curing case.
In other words, a reduction of the material properties of
the pre-reinforced zone makes it possible to model the
time-dependent effect of the pre-reinforced zone on the
global tunnel behavior upon initial tunnel excavation.
Figure 30(b) shows the normalized horizontal displacement at the tunnel face. It is found
that the normalized horizontal displacement for the
time-dependent condition varies within a range of
0.94 0.98, which is very similar to that of other cases
during excavation. Therefore, pre-reinforcement can
be considered for prevention of collapse rather than
as a means of displacement reduction control at the
tunnel face. Thus, it can be concluded that grouting
reduces the horizontal displacement by approximately
2 6% at the tunnel face with the pre-reinforcement
method.
57
Table 7.
Summary of equivalent design methods.
Figure 31. Diagram of simplified pre-reinforced zone.
An analysis method combining experimental

and numerical procedures that consider the timedependent effect on the pre-reinforced zone on tunnel
behavior will provide a reliable and practical design
basis and means of analysis for tunnels in soft ground.
5.3 Determination of equivalent design parameters
for the pre-reinforced zone in residual soil
Table 8. Equivalent design properties used for numerical

modeling.
The design and analysis of pre-reinforcement techniques require design assumptions that are problematic
at best, resulting in increased uncertainty in tunnel
design. The pre-reinforcement effect is typically modeled by simulating the construction sequence of setting
the reinforced zone and then increasing the stiffness,
thereby obviating the need for complex modeling of
each bulb and steel pipe. However, this approach has
been found to be unsuitable when the center-to-center
distance between pipes is larger than the expansion of
the grout bulb. This approach assumes that the stiffness
of the pre-reinforced zone is the same as the stiffness
of the grout bulb, which may be either completely
hardened or arbitrary, and that the pre-reinforced
zone becomes two to four times stronger than before
reinforcement. Therefore, this study presents a new
technique for determining a reasonable equivalent
parameter of the pre-reinforced zone.
A pre-reinforced zone consists of ground, grout
bulbs, and steel pipes. It may be simplified, as shown in
Figure 31, to a condition where the space between the
pipes is wider than the grout expansion range. In sandy
soil and weathered soil, the bulb may be cylindrical.
We assessed five cases comprising various conditions
to model the pre-reinforced zone, as summarized in
Table 7.
For the strengthening of the grout bulb reinforcement, we followed the method of Kikuchi et al. (1995).
The grout injection consequently produced a tenfold
increase in the stiffness of the weathered soil. Table 8
shows the equivalent design parameters for the various compositions of the ground, the bulb, and the
steel pipes. A precisely-simulated model (Figure 32a)
Equivalent design properties
Eeq (MPa)
Ceq (kPa)
Case 1
Case 2
Case 3
Case 4
Case 5
1748.21
107.35
117.81
315.22
490.35
1065.04
91.21
141.22
216.73
588.42
is compared with various equivalent stiffness models

(Figure 32b).
Figure 33 shows the vertical displacement at the
tunnel crown, the horizontal displacement at the
springline, and ground surface settlement. To obtain
these values, we averaged the five nodes located at the
center and left areas of the tunnel crown at a depth of
4.5 m from the portal. The DRM/DEM parameter is
herein defined as the fraction of the displacement of
the referential model (DRM) to the displacement of
the equivalent model (DEM):
The precisely-simulated model is represented by a

value of 0%.
From the vertical displacement results, case 1 and
case 2 give the closest result to the precisely-simulated
model in weathered soil. However, general methods
tend to overestimate the effect of pre-reinforcement.
Although there is only a slight difference between
Case 2 and Case 3, Case 2 predicts a similar
horizontal displacement at the springline relative to the
58
Figure 32. 3D FE analysis model for comparison of equivalent model and precisely modeled.
precisely-simulated model and the design is rather

safe. The equivalent model offers a satisfactory prediction of the ground surface settlement, as the
DRM/DEM value ranges from 0.55% to 0.7% in Case
1 and Case 2.
Case 1 exemplifies a proper equivalent modeling
technique for simulating the pre-reinforcing mechanism in residual soils. Case 2, a SPSS in which the
stiffness of the bulb and steel pipes are coupled in parallel and then connected to the stiffness of the ground
in series, exemplifies a simple equivalent modeling
technique that predicts the vertical displacement at
the tunnel crown, the horizontal displacement at the
springline, and the ground surface settlement.
When the ground, grout bulbs, and steel pipes are
regarded as an individual support system, the prereinforced zone is not a series or parallel stiffness
system but a series-parallel compound stiffness system. Thus, a small degree of stiffness support receives
the largest stress; moreover, a linear combination of
large stiffness supports resists ground displacement.
The SPSS explains the failure mechanism of the
Figure 33. Comparison between equivalent cases.
pre-reinforced zone: namely, the rupture of the steel

pipes and the grout bulb follows the yielding of the
ground. The SPSS should be used in various studies
related to the analysis of pre-reinforced tunnels.
59
EFFECT OF SPATIAL VARIABILITY
The mean value of the measurements is often used

for design parameters even if there is a noticeable
variation. The effect of the variation of geotechnical
parameters on tunnel safety or deformation is rarely
studied or considered using a statistical concept. There
are two kinds of sources for variation in the design
parameters: spatial distribution and uncertainty.
Uncertainty means that the material property of
the soil/rock has a characteristic unreliability. In
reliability-based designs, the uncertainty of the geomaterial is significant for a specific site characterization. There are three major geotechnical uncertainties governing the variability of geoproperties:
inherent soil characteristics, measurement errors, and
transformation fallacies (Phoon & Kulhawy, 1999).
The spatial distribution of soil has been considered for ecological and environmental modeling (Jury,
1985) and various characterization methods have been
suggested by Cho et al. (2004). Likewise, the spatial distribution of the geoproperties is important
for the mechanical behavior of underground structures surrounded by variable soils. The soil itself
is not an isotropic material, but an anisotropic and
non-homogenous material. These characteristics are
induced by the formation process and ground stress.
Spatial distribution takes the macro scale into the range
of interest. Thus, it is expected that, among other soils,
weathered residual soils have high spatial variability of
their geoproperties due to their origin and weathering
process.
The effect of the spatial distribution on the geotechnical parameters of tunnel deformation is studied
through numerical analyses based on statistical concepts as shown in Figure 34. The geotechnical parameters that cause the largest deformation of tunnels
when the ground material follows the Mohr-Coulomb
model and the expected displacement variation characteristics for each geotechnical design parameter are
presented in this study.
The coefficient of variation (COV) has long been
commonly used to quantify the variability of soil and
rock properties (Harr, 1987). The COV is defined as
the standard deviation () divided by the mean () of
the parameter:
Figure 34. Numerical model of spatial distribution of geoproperty (cohesion, Range = 3R, COV = 40%).
Table 9. Representative coefficient of variation (COV) of
the geotechnical parameters.
Properties
COV(%)
Reference
Elastic Modulus
15 45
Friction Angle
Cohesion
24 32
40 68
Harr, 1987; Phoon and

Kul-hawy,1999
Schultze, 1972
Schultze, 1972; Tan
et al. 2000
Tang, 1997; Phoon & Kulhawy, 1999; Tanit et al.,

2004). Thus, it is assumed herein that all geotechnical parameters, i.e., constitutive components of the
Mohr-Coulomb model, have normal distribution characteristics.The representative COV of the geotechnical
parameters is summarized in Table 9.
The bulk modulus (B) and shear modulus (G) are
important material parameters and can be alternatively
changed by the elastic modulus (E) and Poissons ratio,
respectively. The normally distributed elastic modulus
generated with a specific COV is converted to a bulk
modulus and shear modulus to simulate the spatial distribution effect of the elastic modulus on the tunnel
deformation.
The numerical analysis results are shown in Figure
35(a). A COV of 20% should be considered a critical
variation of the elastic modulus effect on the tunnel
deformation, as a COV of more than 30% causes a relatively significant deformation variation. The elastic
modulus is an influential geoproperty in tunnel deformation. The largest NDVR is more than 0.075 when
COV = 40% and range = 3R. The displacement variation induced by the elastic modulus variation can be
predicted from Figure 35(a), depending on the range
from the tunnel center.
Each material property of the residual soil can be

regarded as a normal random variable that has a certain probabilistic error. In particular, it has been shown
that the spatial distribution of the friction angle and
unit weight follows a normal distribution parameterized with the mean and COV (Lumb, 1966; Hoeg
& Murarka, 1974; Lacasse & Nadim, 1996; Low &
60
for the calculation of deformation in the tunnel compared with other geotechnical parameters when the
Mohr-Coulomb material model is used. The general
COV of the friction angle is small at 12% (Schultze,
1972). Therefore, a range of 10%20% variation of
the friction angle is critical for tunnel deformation
characteristics. It can be seen that as the variation range
increases, the slope of the curvature increases.
The COV of cohesion is near 40% (Fredlund &
Dahlman, 1972). As shown in Figure 35(c), the NDVR
enlarges in all cases as the COV increases. However,
it can be observed that the variation of cohesion has a
minor effect on deformation in the tunnel. The maximum NDVR induced by the cohesion variation is lower
than 0.00006 when the variation range is 3R and the
COV is 40%. From the analysis results, the spatial distribution effect of cohesion on tunnel deformation is
smaller than that of the elastic modulus and friction
angle. Therefore, the spatial distribution of cohesion
is not a critical parameter for the characterization of
tunnel behavior.
7
CONCLUSIONS
We presented an overview of geotechnical aspects of

underground construction in urban areas where mostly
decomposed residual soils are present, focusing on
mechanical properties, apparent earth pressure, effect
of ground water, and effect of spatial variability in
geotechnical properties.
The strength and deformation characteristics of
residual soils are affected by particle/pore size, fines
content, mineralogy, and unsaturation, among others:
The internal friction angle decreases with an increase
of fine content and partial saturation increases the
strength of unsaturated residual soils due to capillarity.
Results of back analyses from actual measurements
show that geotechnical properties of residual soils,
which are commonly used for design and modeling
in Korea, are inappropriate in this regard, thus demonstrating the importance of the observational method in
tunnel engineering.
As an unsaturated soil is re-saturated, its apparent cohesion can be eliminated. Thus, during tunnel
construction, cohesion loss can be induced by resaturation (e.g., seepage hindrance, drainage clogging
and groundwater change) and may result in tunnel face
instability.
The lateral earth pressure coefficient of residual
soils is smaller than the coefficient commonly suggested in the literature (e.g., 70 80% of Pecks
suggestion).
Unexpected groundwater inflow and seepage forces
often cause tunnel failures during construction.
Several case histories suggest that tunnel face collapses always occur along with seepage ahead of the
tunnel face. Analytical results suggest that the total
Figure 35. Effect of spatially distributed geoproperties on

tunnel deformation.
The effect of spatial distribution in the friction angle

on tunnel deformation is shown in Figure 35(b). The
NDVR at the tunnel crown and springline reaches
0.0770.078 when the variation range is 3R and the
COV of the friction angle is 40%. Therefore, the friction angle variability is the most high-ranking factor
61
Numerical and Analytical Method in Geomechanics 30:

137155
Fredlund, D.G. & Dahlman, A.E. 1972. Statistical geotechnical properties of glacial lake edmonton sediments, in
Statistics and Probability in Civil Engineering. Hong
Kong University Press, distributed by Oxford University
Press. London.
Harr, M.E. 1987. Reliability Based Design in Civil Engineering, Dover Publications, INC.
Hoeg, H. & Murarka, R.P. 1974. Probabilistic analysis
and design of a retaining wall. Journal of Geotechnical
Engineering Division. ASCE 100 (3): 349366
Jury, W.A. 1985. Spatial variability of soil properties.In:
Hern, S.C., Melancon, S.M. (Eds.), Vadose Zone Modeling
of Organic Pollutants, Lewis, Chelsea, MI, 245269
Kikuchi, K., Mito, Y. & Adachi, T. 1995. Case study on
the mechanical improvement of rock masses by grouting.
Rock Foundation. Balkema 393398
Kim, J.S., Choi,Y.K., Park, J.H., Woo, S.B. & Lee, I.M. 2007.
Effect of viscosity and clogging on grout penetration characteristics. Journal of Korean Geotechnical Society 23 (4):
513
Kim, Y.J. 1993. Mechanical characteristics of decomposed
Korean granites. Ph.D. Thesis, Seoul: Korea University
Lacasse, S. & Nadim, F. 1996. Uncertainties in characterizing soil properties.Uncertainty 96, Geotechnical Special
Publication, ASCE 58 (1): 4975
frictional material. Geotechnique 40 (4): 581606
Lee, C.K. & Jeon, S.K. 1993. earth pressure distribution on
retaining walls in the excavation of multi-layered ground,
Journal of Korean Geotechnical Society 9 (1): 59
Lee, I.K. 1991. Mechanical behavior of compacted decomposed granite soil. Ph.D. Thesis, London: City University.
Lee, I.M., Moon, H.P., Lee, D.S., Kim, K.R. & Cho, M.S.
2007. Earth pressure of vertical shaft considering arching effect in layered soils. Journal of Korean Tunnelling
Association 9 (1): 4962
Lee, I.M., Nam, S.W. & Ahn, J.H. 2003. Effect of seepage
forces on tunnel face stability, Canadian Geotechnical
Journal 40: 342350
Lee, I.M., Park, Y.J. & Reddi, L.N. 2002. Particle transport
characteristics and filtration of granitic residual soils from
the Korean peninsula. Canadian Geotechnical Journal 39:
472482
Lee, I.M., Sung, S.G. & Cho, G.C. 2005. Effect of stress state
on the unsaturated shear strength of a weathered granite.,
Canadian Geotechnical Journal 42: 624631
Lee, S.J., Lee, S.R. & Jang, B.S. 2002. Unsaturated shear
strength characteristics of weathered granite soils. Journal
of Korean Geotechnical Society 22 (1): 8188
Low, B.K. & Tang, W.H. 1997. Reliability analysis of reinforced embankments on soft ground. Canadian Geotechnical Journal 34 (5): 672685
Lumb, P. 1966. The variability of natural soils. Canadian
Geotechnical Journal 3 (2): 7497
Peck, R.B. 1969, Deep excavations and tunneling in soft
ground, Proc., 7th Int. Conf. on Soil Mech. And Found.
Engrg., State-of-the-Art Rep., State-of-the-Art Vol:
225290
Phoon, K.K. & Kulhawy, F.H. 1999. Characterization of
geotechnical variability. Canadian Geotechnical Journal
36 (4): 612624
support pressure is little affected by the tunnel depth

and increases significantly with an increase in the
groundwater level ratio. As the total support pressure
is related to the tunnel face stability, the seepage force
seriously affects the tunnel face stability.
Particle transport characteristics of residual soils
might be among the factors that result in the instability
of underground structures. Their phenomena are complicated and are involved in erosion versus self-healing
(redeposition) processes.
While groutability is affected by the pore size of the
residual soils, proper selection of a grouting method
has proved a difficult task in granite residual soils.
When seepage problems are anticipated during tunnel construction, proper grouting around tunnels can
achieve effective reduction of seepage force acting on
the shotcrete lining and can increase the stiffness and
strength of the surrounding ground.
An analysis method combining experimental
and numerical procedures that consider the timedependent effect on the pre-reinforced zone on tunnel
behavior will provide a reliable and practical de-sign
basis and means of analysis for tunnels in soft ground.
It is expected that, among other soils, weathered
residual soils have high spatial variability of geoproperties, because of their origin and weathering process.
The numerical results show that tunnel deformation
increases with an increase in the spatial variability of
geotechnical design parameters and is accelerated with
an increase in tunnel size.
Research on decomposed residual soils has been
conducted throughout the last few decades. Nonetheless, the current increase of underground construction
projects in urban areas requires better understanding
of residual soils for safer underground use.
ACKNOWLEDGEMENTS
This study was funded by the Korea Institute of Construction and Transportation Technology Evaluation
and Planning under the Ministry of Construction and
Transportation in Korea (Grant No. 04-C01).
REFERENCES
Barisone, G., Pigorini, B. & Pelizza, S. 1982. Umbrella arch
method for tunnelling in difficult conditions Analysis
of Italian cases. Proceedings of the 4th Congress International Association of Engineering Geology. New Delhi, 4:
1527
Cho, G.C., Lee, J.S. & Santamarina, J.C. 2004. Spatial variability in soils: High resolution assessment with electrical
needle probe. Journal of Geotechnical and Geoenvironmental Engineering. ASCE 130 (8): 843849
Cho, K.H., Choi, M.K., Nam, S.W. & Lee, I.M. 2006.
Geotechnical parameter estimation in tunneling using relative convergence measurement. International Journal for
62
Schultze, E. 1972. Frequency distributions and correlations

of soil properties. In Statistics and Probability in Civil
Engineering, Hong Kong University Press, distributed by
Oxford University Press, London.
Shin, J.H., Lee, I.K., Lee, Y.H. & Shin, H.S. 2006. Lessons
from serial tunnel collapses during construction of the
Seoul Subway Line 5. Tunnelling and Underground Space
Technology 21: 296297
Shin,Y.J. 2007. Elasto-plastic ground response of underwater
tunnels considering seepage forces, Ph.D. Thesis, Seoul:
Korea University
Song, K.I., Kim, J. & Cho, G.C. 2007. Numerical analysis of pre-reinforced zones in tunnel considering the
time-dependent grouting performance. Journal of Korean
Tunnelling Association 9 (2): 109120
Tanit, C. 2004. Reliability-based design for Internal stability of mechanically stabilized earth walls. Journal of
Geotechnical and Geoenvironmental Engineering. ASCE
130 (2): 163173
Yoo, C. 2001. Behavior of braced and anchored walls in soils
overlying rock. Journal of Geotechnical and Geoenvironmental Engineering. ASCE 127 (3): 225233
63
General reports
Safety issues, risk analysis, hazard management and control

C.T. Chin & H.C. Chao
Moh and Associates, Inc., Taipei, Taiwan
ABSTRACT: The occurrence of hazard events in geotechnical practice is often associated with geotechnical
uncertainties. To prevent hazard events from happening or reduce the impacts of their consequence, sources of
geotechnical uncertainties need to be identified and treated, and appropriate control measure has to be implemented throughout a project. The probabilistic methods are the devices dealing with uncertainties and the risk
management is the technique to facilitate achieving the goal of project safety. The papers presented in this theme
are summarized and briefly discussed in three categories risk and decision analysis, geotechnical control, and
analysis and control of ground response. In risk analysis, geotechnical uncertainties need to be treated explicitly.
Results of risk analysis, qualitatively or quantitatively, not only provide a baseline for decision making but also
insights to the problem of concern. In order to make best use of available analysis tools, more case studies are
needed.
1
1.1
INTRODUCTION
Geotechnical uncertainty
Uncertainty is the lack of certainty, a state of having

limited knowledge where it is impossible to exactly
describe existing state or future outcome (Hubbard,
2007). If uncertainty requires to be treated, probability is the device. There are two kinds of probability:
relative frequency and subjective, degree-of-belief
probability. The probability of an uncertain event is the
relative frequency of occurrence when it is obtained
through repeated trials or experimental sampling. The
subjective, degree-of-belief probability comes from
judgment where the probability of an uncertain event
is the quantified measure of ones belief or confidence,
according to the information available and ones state
of knowledge at the time it is assessed (Vick, 2002).
In engineering practice, although uncertainty is
surely a certainty and may create some impact, it seems
the existence of which is not a bother to many. As a
matter of fact, most of the engineers get along quite
well without explicitly using the necessary device,
probability, to manage it. This probably attributes to
the customary practice engineers have. By using the
established standards, codes, factors of safety, design
criteria, or procedures in which the uncertainties somehow have been accounted for or somewhere hidden
behind, deterministic method by which answers are
either correct or wrong can comfortably be applied in
solving problems.
However, the material in geotechnical engineering is widely known for its significantly abundant
Figure 1. Categories of uncertainty in soil property data

(Christian et al. 1994).
uncertainties. Sources of uncertainties can be

unknown presence of geologic defects, uncertain value
of soil properties, limited knowledge to the mechanisms and processes, and much more. One of the
examples as presented by Christian el al. (1994) is
demonstrated in Figure 1, in which the uncertainties
associated with the characterization of soil properties are attribute to systematic error and data scatter.
Systematic error results from statistical error in mean
value arising from limited numbers of measurements
or bias in measurement procedures like those associated with field permeability test or field van test. Data
scatter, on the other hand, is a result of random testing
errors or actual spatial variation in the soil profile.
Another source of geotechnical uncertainty is modeling. A model is an appropriate simplification of
reality. Good modeling skill is reflected in the ability
to identify the appropriate level of simplification to
recognize what features are important and what are not.
67
Very often engineers are unaware of the simplifications that they have made and problems may arise
precisely because the assumptions that have been made
are inappropriate in a particular application (Wood,
2004).
The types of geomechanical model can be divided
into the following categories: empirical model, theoretical model, analytical model, numerical model and
constitutive model. Model uncertainty is the extent
to which a model incarnates a uniquely correct representation of the physical process it seeks to emulate. Model uncertainties arise from its representative
degree to the real field processes, and for a physical
process different models and operation can always be
found.
In conventional approach, effects of uncertainty
are generally accounted for, consciously or not, by
standards, codes, design criteria, or factors of safety.
The underlying uncertainties may have already been
considered or evaluated somehow in the process developing them. This simple strategy seems work quite
well for works having been encountered before and lot
of experience accumulated. However, when circumstances are unique and uncertainties are not routine,
which are often the cases in geotechnical engineering
practice, aforementioned procedures or tools may not
be fully applicable any more.
Figure 2. Geotechnical safety.
uncertainty in a rational and explicit way. In response

to this demand, active discussion and development
have been seen on topics that may be grouped into
five categories: soil variability characterization, reliability analysis, new generation design code, observational method, and risk assessment and management.
Advancements in these disciplines, on the other hand,
are intended to address better issues on geotechnical
safety. These concepts are conveniently illustrated by
Figure 2.
1.2 Geotechnical safety
1.3 Risk assessment
The occurrence of hazard event in major geotechnical engineering projects such as tunneling or deep
excavation often draws a lot of attention from the general public. Potential consequences include significant
financial loss to the client and contractor, schedule
delay and loss of confidence to the general public, and
casualty. Thus, it is obvious that safety is an important
issue that requires special cares to ensure the objectives
of a project can be achieved.
Safety itself is in essence not a measurable quantity.
In practice, it is evaluated through the safety indicators.
Some of the physical characteristics such as the size of
cracks, deformations and differential settlements, are
selected to serve as safety indicators for evaluating
the safety status. This process requires analysis and
interpretation, and judgment plays an important role
in it. If the uncertainty is to be accounted for and the
effects are accommodated, sources of uncertainty must
be understood first. This is where probability comes
in. One of the most important tools in dealing with
probability is judgment. Judgment provides an interpretative framework that helps guide how uncertainties
are comprehended and subsequently managed. When
a structure is said to be safe during adjacent excavation, it means the assessor hold some sufficient degree
of belief.
In past decades, there has been a greater awareness on need to treat and manage the geotechnical
Traditionally, risks were managed indirectly through

the engineering decisions taken during the project
development. More often than not, the information and
knowledge used behind the decisions are inexplicit
and not easy to trace. To improve this, it is necessary to introduce formal risk management technique.
In this practice, risk is defined as the product of failure
probability and consequences:
Where, Pf is probability of failure and Cf is consequences of failure.

By definition, risk assessment must account for
both the probability and extent of adverse consequences of hazard events arising from a given activity.
Thus, risk assessment involves identification of hazard events and qualitative or quantitative descriptions
of risks. The scope of risk analysis contains the entire
process of the causes and effect of adverse events.
Such a process is composed of three sequential parts,
an initiator that starts it, response to the initiator
and the consequences (Vick, 2002) as illustrated in
Figure 3.
Initiator is the cause that sets a potential failure process in motion. Response is the event directly resulted
from the initiator, and also called the hazard event
in some occasions. Response leads to failure if the
68
Figure 3. Components of risk analysis (Vick, 2002).
resistance or capacity of a structure is unable to sustain the effects. In risk analysis, failure is treated as
consequence and needs to be a physically observable
condition like retaining wall collapse, utility pipeline
broken, roadway damage et al. Generally, failure cannot be evaluated meaningfully if it is not converted
to something measurable such as lives lost or dollar cost. It is also important to realize there are
uncertainties associated with the response. As shown
in Figure 3, probabilistic methods such as Bayes
approach, reliability analysis, subjective judgment and
others are introduced as the tools for evaluating these
uncertainties.
As shown also in Figure 3, initiator, response and
consequence are linked together by decomposition
techniques such as event tree and fault tree. Event tree
and fault tree facilitate envision how failure process
occurs. As such, they are basic tool of risk identification and analysis. Figure 4(a) shows a two-level
of event tree and fault tress. The event starts with
initiator event I. If it occurs, the next event to happen is R1 , followed by R2 or R3 or both. This event
tree contains two failure modes IR1 R2 and IR1 R3 .
Figure 4(b) presented the same failure progress by
Fault tree. Event uses a bottom up or forward
approach beginning with the initiator and taking it
to consequences. This captures failure process by
expressing its logical order. Fault tree, on the other
hand, uses a top down or backward approach starting from the top event and identifying the possible
events in the 2nd level through a search process. The
and and or logic gates instrumented in Fault tree
allow computations using Boolean algebra in coping
with complex problem.
Figure 4. Event tree and Fault tree format.
After risk is identified, the rest of work in risk analysis is to evaluate the probability and consequences
such that the corresponding rank-order can be determined based on predefined criteria. Subjective,
degree-of-belief probability often are assessed using
expert investigation method incorporated with other
techniques such as Dephi method, Analytic Hierarchy
Process (AHP), Fuzzy set theory in risk analysis. One
of the controversial issues arises here in the process
quantification of risks. According to Vick (2002),
qualitative approaches allows us to discern that one
thing is more likely or less likely than some other, but
an important property by how much more or less in
the qualification process is missing. Other than it, the
depth of the insight to the problem can be reached is
considered in proportion to the efforts made in the process of risk analysis where numerical quantification
usually requires more.
Risk response measure and reduction strategy can
be developed based on risk acceptable criteria defined
in risk policy (ITA, 2004). For risks with high rankorder, it may be necessary to develop design or alternatives in determining risk reduction measure. Decision
analysis is essentially a comparative approach based
on decision rules and results of risk analysis on various alternatives. Similar techniques in risk analysis
69
Probability Density Function
Table 1.
Risk and decision analysis.
ID
Topic
IS-022
Risk analysis and fuzzy comprehensive

assessment on construction of shield
tunnel in Shanghai metro line
Risk assessment on environmental impact in
Xizang Road Tunnel
Risk analysis for cutterhead failure of composite
EPB shield based on fuzzy fault tree
Risk assessment for the safe grade of deep
excavation
Multi-factors durability evaluation in subway
concrete structure
Research on structural status of operating tunnel
of metro in Shanghai and treatment ideas
FS=1.50
Design alternative 1:
low uncertainty
IS-294
FS=2.00
(Pf)2=7.4%
0
1
(Pf)1=0.6%
Design alternative 2:
high uncertainty
IS-293
IS-070
Factor of safety, FS
IS-055
Figure 5. Effect of uncertainty.
IS-383
can be applied to evaluate the risk associated with

each of the alternative. By comparing the results of
risk analysis, the alternative having the greatest risk
effectiveness can be identified. Results of decision
analysis provide a solid baseline for decision maker
to make decision when dealing with risk.
1.4
REVIEW OF PAPERS
2.1 Risk and decision analysis

As listed in Table 1, the papers related to risk and
decision analysis are grouped into this section. Among
these papers, IS-022, IS-294 and IS-293 discuss tunneling risk analysis, and IS-070 introduces risk analysis for deep excavation, and IS-055 is about decision
analysis. While the others apply formal risk or decision analysis techniques in the evaluation approach,
the approach presented in paper IS-383 can be taken
as an informal risk management approach. As have
been used widely in geotechnical practice, risks associated with the identified hazard events were not
presented. Nevertheless, the causes of hazard events
were identified and the hazard prevention measures
were developed.
Paper IS-022 presented a risk analysis approach in
which risks were identified thorough the work breakdown and fault tree method. The associated rank-order
of the identified risks were analyzed by fuzzy comprehensive evaluation method. The expert investigation
method and AHP were used to determine the risk
indicators weight, evaluate the probabilities and consequences of the identified risks. The membership
function was applied to determine the membership
degree value of each of the risk events. The case investigated in this paper was Shanghai metro, which is
a 32.2 km long shield tunnel. For the convenience of
risk analysis, the metro line was divided into 12 sections. Results of the evaluation showed the identified
risk events include obstruction, tunnel collapse, quicksand, groundwater ingress, ground settlement, et al.
The risk levels with respect to each of the sections
were from medium to significant. The possible financial loss is between 100,000 RMB to 10,000,000 RMB.
Based on the risk classification criteria of this case, the
consequences are great but compensable.
Deterministic thinking versus probabilistic

thinking
In engineering, a deterministic system is a system that

given a particular input, it will always produce the same
output, and the underlying machine will always pass
through the same sequence of states. In this system,
results of thinking are either right or wrong, mathematic algorithms used are either correct or not, and
there is no need to be bothered by uncertainty.
However, uncertainty in fact exists and the effects of
which may be significant in many occasions and cannot just be neglected. As has been mentioned earlier in
this report, uncertainties in geotechnical engineering
practice are managed through standards, codes, design
criteria, factors of safety or established procedures.
With such strategies, the deterministic thinking works
and functions well for situations where experience
and routine works dominate. For unique situations or
non-routine conditions or in risk analysis, uncertainty
becomes an unavoidable issue that requires using of
the technique of probabilistic methods in obtaining
appropriate solution.
Figure 5 demonstrates the effect of uncertainty in
design results for two alternatives based on different design models. The computed factors of safety
for alternative 1 and alternative 2 are 1.50 and 2.00,
respective. In a deterministic system, the factor of
safety for alternative 2 is apparently higher, and this
alternative is almost certain to be selected as the final
scheme. However, if the uncertainties associated with
both models are considered, the failure probability of
alternative 2 is apparently higher even the computed
factor of safety seems better.
70
of the evaluation indicated that the level of safety

of the project as a whole was acceptable. However,
size of excavation area and hydrogeology condition
were identified as the factors having more significant
impact on the safety of the excavation activities. Measures to prevent hazard events from happening were
thus developed based on the results of risk assessment.
Paper IS-055 presented a decision analysis
approach for determining the optimal design alternative for the concrete mixture of a subway tunnel.
Factors having significant influence on the durability of subway concrete structure were identified to
be stray current corrosion, chloride ions ingression,
sulfate attack and carbonation. While the durability
attribution from each of the factors was determined
individually, the combined influence of these factors
was vague. To obtain the optimal design scheme, five
design alternatives were developed. Expert investigation method, AHP and Multiple Attribute Decision
Model incorporated with fuzzy set theory were applied
in assessing the joint influence of these factors. Results
of the evaluation provided a baseline for decision
maker in selecting the adequate design scheme.
Paper IS-383 presented the health diagnostic results
and hazard prevention or reduction measures for an
operating tunnel of Shanghai metro. Results of the
diagnosis showed the tunnel structure is suffered from
detrimental events such as ingress of groundwater,
crack and convergence of tunnel lining, and settlement
along the tunnel alignment. The causes of these detrimental events were identified to be adverse geological
condition, sequela of the accidents during construction, local ground subsidence, construction activities
in the proximity, etc. To prevent these detrimental
events from getting worse and reduce the associated
risks to the tunnel and surrounding environment, metro
passengers and third parties, control measures in terms
of regulation, monitoring schemes were proposed.
Paper IS-294 presented the risk analysis for the

construction of Shanghai Xizang Road Tunnel, a rivercrossing tunnel built for the 2010 Shanghai Exposition. Other special features included up-cross and
down-cross existing metro lines. Risk analysis was
based on expert investigation and confidence index
methods. The computer software TRM 1.0 developed by the Tongji University was used as the tool
to analyze the risk data and thereafter determine the
corresponding risk level. In the case investigated, the
work of risk analysis was focused on the environmental impacts as a result of tunnel construction to
the surrounding buildings, roads, and utility pipelines.
Based on the analysis results, risk mitigation measures
focused on ground settlement control were proposed.
The authors also pointed out that the risk is dynamic in
nature, which requires constant and cycled assessment
and treatment.
Paper IS-293 presented for tunneling in adverse
ground condition, the cutterhead of shield machine is
exposed to high risk of failure as a result of heavy
demand in machine operation. This paper presented
a risk analysis approach for cutterhead failure using
the fault tree analysis method and fuzzy set theory.
In this study, risk identification was conducted by
the fault tree analysis method. Results of the identification indicated three major hazard events cuter
disk failure, cutter failure and other system components failure will lead to cutterhead failure. Expert
investigation method and fuzzy set theory were introduced to evaluate the probabilities and consequences
of the occurrences for these hazard events. Based on
the results of risk analysis, remediation measures for
the hazard events with higher risk level were developed
and presented.
Historic data in paper IS-070 showed the proportion
of accidents in deep excavation projects resulted from
design and construction related problems is 87% based
on 344 cases investigated. Because effective tools in
dealing with geotechnical uncertainty and applicability of current analytical theories are limited, engineers
tend to be conservative consciously or not in dealing
with geotechnical problems. This is usually reflected
in high construction costs. To improve this, a Fuzzy
synthetic evaluation process was used in this study
for risk analysis. This method follows the procedures
in which the risk factors were identified first, and
expert investigation method, Analytic Hierarchy Process (AHP) method and Delphi method were adopted
to determine the weight for each of the factors and the
safety index, and the safety ranking order of an excavation project can thus be decided. The case visited was
the Shanghai international passenger transport center
project. The size of the excavation area, hydrogeology condition, design feature, construction aspect and
surrounding environment are selected as the primary
factors affecting the safety of deep excavation. Results
2.2
Geotechnical control
The geotechnical control is a process exercised

throughout the planning, design and construction
phases of a project to facilitate achieving project objectives. The control measures applied include regular
audit inspection, site supervision and risk management to ensure adequate design standards and effective
safety protection be performed such that the project
can be completed in a manner of optimization. As
listed in Table 2, two of the papers were grouped into
this section. The paper IS-372 introduced a metro railway under construction and the paper IS-380 presented
a retrospect study of a cable duct crossing project.
Both of the cases reported were construction projects
in Hong Kong.
Paper IS-372 presented a geotechnical control
process exercised for a HK$8.3 billion subway
71
Table 2.
ID
Topic
IS-372
Geotechnical control of a major railway project

involving tunnel works in Hong Kong
Performance Review of a Pipe Jacking Project
in Hong Kong
IS-380
level management. Specified responses corresponding

to each level of the system were also defined before the
construction started. During the construction, a close
to action level reading on the track settlement was
recorded. Under the framework of geotechnical control process, an urgent meeting was held among the
client, design and contractor. After the meeting, the
type of lubricant used for filling the gap between
the tunnel lining and surrounding soils was replaced.
With the control process, the maximum settlement of
railway tracks was controlled within the maximum
allowable range throughout the whole construction
phase.
Geotechnical control.
construction project in Hong Kong. This project is

packed into three design-built contracts.The content of
the project includes the construction of shield tunnels,
cut-and-cover tunnels and underground station. The
Geotechnical Engineering Office (GEO), Civil Engineering and Development Department of the Hong
Kong Special Administration Region (HKSAR) is in
charge of controlling the building ordinance and regulations and issuing technical standards. Under of
auditing of GEO, the private owner of this project, the
Kowloon Canton Railway corporation (KCRC) committed to follow the Joint Code of Practice for Risk
Management of Tunneling Works for implementation
of risk management process. The geotechnical control process started from the planning stage and will
be kept in effect throughout the construction stage.
Because the project is considered a private project,
the tunnel works may be exempted form the administrative procedures. Thus, a review panel was formed
within GEO in the planning stage. The KCRC demonstrated to GEO they had met the requirements for
instrument of exemption (IoE) in aspects including
risk management, design, and construction. The IoE
was issued after the document met with the specified
requirements. Under the IoE, the KCRC is required
to appoint authorized personnel, employ assurance
system and control scheme, prepare site supervision plan, and keep appropriate records and reports
for regular GEO inspection during the construction
stage. Monthly meeting are scheduled between relevant parties. When significant changes occur in design
or working methods, the KCRC needs to report to
GEO. The geotechnical control process is currently
in progress.
Paper IS-380 review the performance of a cable
duct crossing construction project in Hong Kong. With
the project, the geotechnical control process was exercised by the GEO, HKSAR. This project involved
the construction of a 222 m long and 1.95 m diameter tunnel to serve as cable duct crossing a highway,
two MRT tunnels, and two Airport express link tunnels. This cable duct was constructed with pipe jacking
method. Prior to the commencement of works, a condition survey was conducted for existing utility pipeline.
During construction, an automatic monitoring system
was used to take readings for monitoring railway track
settlement. The work of monitoring was controlled by
a system based on alert level, action level and alarm
2.3 Ground response, analysis and design

Most of the engineering designs are composed of two
major components determining the imposed load
and computing the resistance or capacity. Factor of
safety in a design problem is defined as the ratio
between the resistance and load. Because most of
the geotechnical engineering projects are performed
underground, the construction activities involving add
or remove loads to the ground will inevitably result
in stress and strain redistribution. When the resulted
stresses exceed the resistance or capacity of the
ground, failure could occur. However, ground response
before reaching failure state may draw more concern
for underground construction project. In urban area,
underground construction induced ground deformation make the construction itself and the buildings,
utility pipelines and infrastructures adjacent to the construction site exposed to the risk of being damaged.
Results of ground response prediction and monitoring
can be used as a baseline for developing control and
protection schemes.
The papers in this group are further divided into two
sub-groups ground response, analysis and design
for discussion purposes.
2.3.1 Ground response
Table 3 lists the papers related to the issue of ground
response. Paper IS-048 introduced a prototype and
laboratory scale non-destructive scanning technique
designed for detecting crack or cavity ahead of the
front end of a shield machine. Paper IS-369 presented
an analytical approach for evaluating the squeezing potential of soft rock. Paper IS-247 presents a
numerical investigation for the floor heave behavior
at the T-section of a deep mining tunnel using threedimensional finite element method. Paper IS-376 discusses the application of strain gauge in measuring
strut load in deep excavation project in Singapore.
Suggestions for maximizing its effectiveness are also
proposed. Paper IS-014 and IS-339 present the application of fuzzy set theory and neural network in
predicting ground settlement induced by tunneling.
72
Table 3.
ID
Topic
IS-048
Experimental studies of a geological measuring

system for tunnel with ultrasonic transducer
Squeezing potential of tunnels in clays and
clayshales from normalized undrained shear
strength, unconfined compressive strength
and seismic velocity
Floor heave behavior and control of roadway
intersection in deep mine
Maximising the Potential of Strain Gauges:
A Singapore Perspective
Prediction of surface settlement induced by
shield tunneling: an ANFIS model
The use of artificial neural networks to predict
ground movements caused by tunneling
IS-369
IS-247
IS-376
IS-014
IS-339
For large cross section tunnel, the ground deformation resulted from excavation work is usually much
more significant at the tunnel intersection than at the
regular part. Thus, special measures controlling possible ground movement that might be detrimental to
the safety is necessary. Paper IS-247 presented an
investigation of the floor heave at a T-section of a ventilation tunnel at the depth of GL. 990 m of Tongkou
colliery using three-dimensional finite element modeling. Based on the evaluation results, suggestions for
controlling floor heave were given.
Instrumentation plays an important role in ensuring that construction control is maintained during
excavation. Comparison between the monitored data
and design predictions provides the opportunities for
verifying design results and refining design methods. Experiences indicate in most of the excavation
project, some of the instrumentation readings are genuine load conditions while some are not. It is important
to gain clear understanding of the impact of construction activities and data interpretation and make effort
to maximize the quality of the data.
According to paper IS-376, strain gauge is widely
used in measuring the change of strut load in deep
excavation in Singapore because of the acceptable reliability and low costs compared to other instruments.
However, the performance of strain gauge can easily
be affected by the location where it is installed, electromagnetic interference, temperature, preloading, welding and other construction activities. To minimize the
influence of these factors, measures such as use of
load cell as a cross reference, protection against construction induced disturbance, and use of real time
system were proposed. For data interpretation, skilled
personnel fully aware of the design predictions for the
excavation, the excavation process and the potential
impact of the excavation on the readings should be
assigned.
Due to the complexity and variation of the ground
composition, accurate prediction of tunneling induced
ground settlement based on conventional geotechnical approaches such as empirical methods, analytical
methods or numerical methods is considered by many
as a challenge. Empirical approach is easy to use but
often precludes the consideration of ground resistance
and deformability parameters. Analytical and numerical approach are generally using simplified ground
parameters for practical purposes rather than incorporating the complete ground conditions along the
whole range of a tunnel alignment into the modeling
approach. The applicability of available constitutive
models and simulation of three dimensional ground
response within 2 dimensional space for ground settlement prediction is controversial to many. An alternative in solving the problems mentioned above is the
artificial intelligence technique. Artificial neural network and fuzzy logic were introduced into this field
for prediction in recent years.
Ground response.
Paper IS-048 presented an experimental model

developed for detecting multiple reflection sources
based on rotational scanning technique and ultrasonic
wave reflection method. In the proposed test configuration, the ground was simulated by plaster blocks,
among which the maximum dimension is of 1000 mm
in height, 1200 mm in width and 150 mm in thickness. Horizontal and inclined cracks are simulated by
drilled holes. Various signal process techniques such
as stacking, signal compensation, and demodulation
were applied to obtain ultrasonic image of the space
ahead of the scanning transducer. Results of the experimental evaluation showed that the proposed method
is able to identify the location of multiple reflection
sources.
The tunnel squeezing phenomenon was first
described by Terzaghi (1946) who associated squeezing mainly with clayrich rocks. One of the first
stability criteria to predict squeezing was developed
by Peck (1969) for tunnels in clays based on Broms
and Bennermarks (1967) stability criteria for open
excavation. For tunnel in rocks, most of the squeezing criteria proposed are empirical or semi-empirical
such as Singh et al. (1992), Goel et al. (1995, 2000),
Jethwa el al. (2000) and more. The main challenge in
use of these semi-empirical approaches is in the determination of the rock mass strength. In addition, most
of the proposed methodologies developed thus far are
mainly for clays or hard rocks. Few studies have been
proposed for intermediate material such as hard soils
or soft rocks.
In paper IS-369, simple methods to evaluate the
squeezing potential of intermediate soil-rock material
based on undrain shear strength, unconfined compressive strength and P-wave velocity was proposed.
Evaluation of field measurement and other empirical tunnel squeezing criteria was also performed for
comparative purpose.
73
Figure 6. Conceptual model of a neural network with two

inputs and one outputs.
Figure 7. Architecture of ANFIS.

Table 4.
An artificial neural network (ANN), often just

called a neural network (NN), is a mathematical
model based on biological neural networks. A conceptual model is shown in Figure 6. It consists of an
interconnected group of artificial neurons and processes information using a connectionist approach to
computation. In most cases an ANN is an adaptive
system that changes its structure based on external or
internal information that flows through the network
during the learning phase. In more practical terms neural networks are non-linear statistical data modeling
tools. They can be used to model complex relationship between inputs and outputs or to find patterns in
data. One of the most important properties of ANNs is
their ability to learn from environment and to improve
their performance with such learning. The learning
occurs when the NN reaches a generalized solution
for a particular case of problems.
Fuzzy logic is a form of multi-valued logic derived
from fuzzy set theory to deal with reasoning that is
approximate rather than precise. Just as in fuzzy set
theory the set membership values can range between
0 and 1, in fuzzy logic the degree of truth of a statement
can range between 0 and 1 and is not constrained to the
two truth values, true and false, as in classic predicate
logic. When linguistic variables are used, these degrees
may be managed by specific functions.
The Adaptive Neural-Fuzzy Inference System
(ANFIS) is a hybrid intelligent system combining the
ability of a neural network to fuzzy logic. An illustrative architect of ANFIS is given in Figure 7. It uses
a given set of input/output data to construct a fuzzy
inference system within which the membership function parameters are tuned or adjusted using either a
backpropagation algorithm along or in combination
with a least square type of method. This adjustment
allows the fuzzy system to learn from data and thus
it also has the potential in prediction. This paper presented an application of an ANFIS based model in the
prediction of shield tunnel induced settlement.
Paper IS-339 presented an application of ANN in
predicting ground settlement induced by tunnel excavation. The case visited is the Metro-DF in the city of
Brasilia and the tunnel was constructed using the New
Input and output parameters.
Input parameters, xi
Output parameter, y
Ground settlements along the

centerline of tunnel alignment
0 m, 5 m, 10 m, 20 m and
30 m relative to front end of
shield machine
tunnel lining installed per day in
terms of the number of rings
Ground settlement
along the centerline
of tunnel alignment
+5 m relative to
front end of shield
machine
Austrian Tunneling Method (NATM). Two data sets

were used for training the ANN. One of the data set
was obtained from the monitored data during the tunnel construction and the other was established based
on results of finite element modeling. Validating or
testing of the ANN was performed after the completion of training phase. Results of the evaluation with
monitored data show the average correlation coefficients between the predicted and measured value were
about 0.99 and 0.95 at the completion of training phase
and at the validating phase, respectively. Although
the prediction with finite element method exhibited
better correlation between the results obtained from
the training and validating phases, precision was not
always achieved. The authors attributed the cause of
lacking precision to inadequate constitutive model,
difficulty in simulating the real tunnel geometry, and
simulate the three dimensional physical reality in two
dimensional space. Based on the evaluation results,
the authors concluded that the ANN is an effective
computational tool in predicting settlement induced by
tunneling when good set of training data is available.
Paper IS-014 presented the application of an
ANFIS model in the prediction of ground settlement
induced by shield tunneling. The case visited is the
Shanghai No. 2 subway project. Data set obtained from
field measurements was used in model training and
validating.
Results of the evaluation showed that the ANFIS
predictions are in good agreement with the measured
data with relative error within the range from 2%
to 7%. For this particular case visited, the author
74
depth. The paper IS-125 presents the results of investigation for projects with excavation depth from 14.92 m
to 17.28 m. The comparison of the strut loads between
the design value and field measurement in Shanghai
was also presented. Causes of the deviation were identified to be results of different computation tools, analysis parameters, and lacking communication between
designer and contractor. Over-design and unnecessary cost were often the consequences. Suggestions
on the issues relating to active earth pressure coefficients, vertical spring coefficient, kv , under the toe of
the diaphragm wall, equivalent subgrade coefficient,
kh , strut load, diaphragm thickness, and monitoring
scheme were proposed.
Table 5. Analysis and design.

ID
Topic
IS-083
Research and application of road tunnel structure

optimization
Framework of performance-based fire protection
design method for road tunnel
Discussion on design method for retaining
structures of metro station deep excavation
in Shanghai
Research on stochastic seismic analysis of
underground pipeline based on physical
earthquake model
IS-374
IS-125
IS-234
also compared the ANFIS predicted results with

those derived from Peck approach (1969), Pi-sigma
approach (Gupta and Rao, 1994) and Back Propagation based neural network method. According to
the authors, the ANFIS prediction exhibited relatively
better accuracy and stable in terms of the computed
results.
CONCLUSION AND REMARK
3.1 Treatment for geotechnical uncertainty

In conventional approach, geotechnical uncertainty is
managed through codes, standards, design criteria,
established procedures and other devices. Uncertainty
is recognized somewhat indirectly and inexplicitly.
Engineering judgment plays an important role and
has been used commonly in defining and setting the
aforementioned management devices. For risk management and other unique or non-routine geotechnical
problems, judgment still plays an important role but
geotechnical uncertainty needs to be treated explicitly
for requirements such as safety evaluation. Probabilistic methods including Bayes theorem, reliability
method, subjective probability and more provide the
necessary tools in coping with it.
2.3.2 Analysis and design

The papers relating to analysis and design are grouped
in this section as listed in Table 5.
According to paper IS-083, conventional road tunnels are designed based on passive analysis approach.
This method follows the procedure (1) alternatives
development (2) computation and analysis (3) selecting the best design scheme among alternatives. The
advantage of this method is conceptually easy. However it is time consuming and may not be able to
identify the best alternative if it is not included in
the scheme developed. Thus, it is difficult to evaluate the performance of the structure. To improve
this, an optimization method is proposed in this paper.
In this method, the dimensions of the tunneling are
treated as input variables, and the stress of the spring is
the objective function. Algorithm of complex analysis
programmed by C++ language was used in the optimization process. Results of comparison showed both
the stress distribution along the tunnel and the relating
structural cost were reduced significantly based on the
results evaluation for this method.
Currently, performance based design code has been
used in fire protection design for buildings. However,
the design code for road tunnels is still prescriptive based. In view of this, Paper IS-374 proposed a
framework of performance based design code for road
tunnels with large cross sections.
Although there is no significant variation in the
construction methods and geological condition for
Shanghai metro, results of diaphragm wall and strut
design in terms of diaphragm depth, thickness, rebar
content, and strut load in deep excavation are apparently varied even for projects with similar excavation
3.2 Qualification and quantification for risks

Expert investigation method was used extensively for
risk analysis in determining the probability and consequences of risk events. This method requires experts
to express their judgment in qualitative terms based on
predefined criteria, all information available, and the
uncertainties of the risk they perceive. The application
of elicitation techniques based on cognitive process is
necessary in finding the true belief of the experts. It is
also important to reduce the extent of what considered
unavoidable bias when using elicitation techniques in
expert investigation.
Uncertainty is usually expressed qualitatively and
the use of verbal terms like possible, probable,
likely or unlikely seems to work well. However,
numerical quantification provides the missing link for
qualitative approach. The extent to which the uncertainty of an event is greater than the other can be
determined through quantitative approach. It is important to note that probability, qualitatively or not, is
only a verbal description or the output of numerical
quantification process. The insight obtained through
75
Han, X. & Ding, G.Y. 2008. Framework of performancebased fire protection design method for road tunnel.
Hou, J., Zhang, M.X. & Tu, M. 2008. Prediction of surface
settlement induced by shield tunneling: an ANFIS model.
Kim, D.H., Kim, U.Y., Lee, S.P., Lee, H.Y. & Lee, J.S. 2008.
Experimental studies of a geological measuring system
for tunnel with ultrasonic transducer.
Lam, T.S.K. 2008. Performance review of a pipe jacking
project in Hong Kong.
Lee, W., Chung, S.S., Roberts, K.J. & Pang, P.L.R. 2008.
Geotechnical control of a major railway project involving
tunnel works in Hong Kong.
Li, J.P., Wang, R.L. & Yan, J.Y. 2008. Research on structural status of operating tunnel of metro in Shanghai and
treatment ideas.
Osborne, N.H., Ng, C.C., Chen, D.C., Tan, G.H., Rudi, J. &
Latt, K.M. 2008. Maximising the potential of strain
gauges: A Singapore perspective.
Wang, R., Liu, G.B. & Liu, D.P. 2008. Discussion on design
method for retaining structures of metro station deep
excavation in Shanghai.
Yan, Y.R., Huang, H.W. & Hu, Q.F. 2008. Risk analysis for
cutterhead failure of composite EPB shield based on fuzzy
fault tree.
Yao, C.P., Huang, H.W. & Hu, Q.F. 2008. Risk assessment on
environmental impact in Xizang Road Tunnel.
Zhou, H.B., Yao, H. & Gao, W.J. 2008. Risk analysis
and Fuzzy comprehensive assessment on construction of
shield tunnel in Shanghai Metro line.
the process is what really matters. The insight helps

us understand the true meaning of the probability
associated with the problem and what behind it.
3.3
Decision analysis
Decision analysis is a technique developed on the

framework of risk analysis and comparative skills.
Through this technique, a relatively better design alternative can be determined. Nevertheless, it should be
noted that decision analysis only provides a baseline
for decision maker. A well-informed decision comes
about by considering risk magnitudes, risk reduction
measures, and feasibility in performing these measures
other than the results of decision analysis.
3.4
Need for more case studies
No new theory in soil mechanics can be accepted for

practical use without ample demonstration by field
observations that it is reasonable accurate under a
variety of conditions (Terzaghi and Peck, 1948). To
interpret the results of observation, it is necessary to
incorporate judgment. Judgment, or subjective probability plays an important role for risk management and
hazard control. The acquisition of judgment relies on
experience and knowledge that can be derived from
field observation or diligent study of published case
histories. Because the opportunity of performing personal observation is limited for most in contemporary
geotechnical practice, study of case histories becomes
the primary source for accessing and developing ones
own experience.
For risk management and hazard control, more case
studies are needed. Geotechnical engineers should
make best use of available tools and present their cases
for further studies.
REFERENCES
ABI/BTS. 2004. A Joint Code of Practice for the Procurement, Design and Construction of Tunnels and Associated
Underground Structures, London: The Association of
British Insurer, The British Tunneling society.
Douglas Hubbard How to Measure Anything: Finding the
Value of Intangibles in Business, John Wiley & Sons,
2007.
David Muir Wood. 2004. Geotechnical Modeling, Spon
Press, New York.
Gupta, M.M. & Rao, D.H. 1994. On the principles of fuzzy
neural network. Fuzzy Sets and Systems, 61(1): 18.
Huang, H.W. et al. 2006. Guidelines of risk Management for
Metro Tunnelling and Underground Engineering Works,
Tongji University.
Internatinal tunnel Association. 2002. Working Group No. 2,
Guidelines for tunneling risk management, Balkema.
Peck, R.B. 1969. Deep excavation and tunneling in soft
ground. Proceedings of the 7th International Conference
on soil Mechanics and Foundation Engineering, Mexico.
Steven G. Vick. 2002. Degrees of Belief-Subjective Probability and Engineering Judgment, ASCE Press, American
Society of Civil Engineers, Reston, Virginia, USA.
Terzaghi, K. & Peck, R. 1948. Soil mechanics in engineering
practice, Wiley, New York.
Whitman, R.V. 2000. Organizing and Evaluating Unceratinty in Geotechnical Engineering, Journal of Geotechnical
and Geoenvironmental Engineering, Vol. 126, No. 7, July,
2000.
PAPER IN THEME 4
Ai, X.Q. & Li, J. 2008. Research on stochastic seismic analysis of underground pipeline based on physical earthquake
model.
Bao, X.H. & Huang, H.W. 2008. Risk assessment for the safe
grade of deep excavation.
Chissolucombe, I., Assis, A.P. & Farisa, M.M. 2008. The use
of artificial neural networks to predict ground movements
caused by tunneling.
Cong, C. & Linde,Y. 2008. Multi-factor durability evaluation
in subway concrete structure.
Ding, W.Q. & Xu, Y. 2008. Research and application of road
tunnel structural optimization.
Guo, B.H. & Lu, T.K. 2008. Floor heave behavior and control
of roadway intersection in deep mine.
Gutierrez, M. & Xia, C. 2008. Squeezing potential of tunnels
in clays and clayshales from normalized undrained shear
strength, unconfined compressive strength and seismic
velocity.
76
Calculation and design methods, and predictive tools

LGCIE, INSA-Lyon, F-69621, France
ABSTRACT: This general report covers 19 papers that are included in session 6 of the symposium, related
to the design or calculation methods and predictive tools for tunneling and deep excavations. For this report,
the papers have been classified in 3 main subjects: i) excavations, ii) tunneling, iii) general papers on design
methods and tools. There are a greater number of papers concerning tunnelling, covering a large numbers of
subjects, subdivided in the following topics: T.B.M. simulation, Ground reaction curve, Longitudinal behaviour
of segmented lining, Settlement troughs, Effect of vibrations.
INTRODUCTION
The content of the papers can be broadly divided

into calculation and design of tunnelling works, excavations, and more general papers. 3 papers deal with
excavation, considering 3 different aspects: basal stability, strut loads, and effect on nearby piles. There
are a greater number of papers concerning tunnelling,
covering a large number of subjects. Among these
12 papers, 1 reports on TBM numerical simulation,
2 present analytical methods for the ground reaction
curve, 2 deal with the assessment of the settlement
trough, 2 consider the problem of the longitudinal
behaviour of the segmented tunnel lining, 3 report on
the effect of vibrations or on the seismic response, the
2 remaining concerning more problems related to rock
tunnels. Finally there are 4 general papers concerning
the simulation tools or presenting a national report.
Considering this large number of subjects, it is not
really possible to highlight a main emphasis, some of
the papers covering very narrow subjects and others
concerning very general topics.
In the following sections of this report, the major
findings and key features of each paper are presented
and briefly discussed.
This session with a very broad theme, contains 19

papers. Tables 1 and 2 present a classification of these
papers, by countries and by themes. There are papers
from 8 countries, but more than half of the papers
(11/19) are from China.
Table 1.
Classification by countries.
Countries
Number of papers
Brazil
China
Japan
Kazakstan
Korea
Netherlands
UK
USA
2
11
1
1
1
1
1
1
Table 2.
Classification by subjects.
Topics
Sub-topic
Excavations
Tunnels
General
TBM simulation
Ground reaction curve
Settlement trough
Longitudinal behaviour
Effect of vibrations
Number
of papers
EXCAVATIONS
Song and Huang studied the basal stability of an excavation in soft clay by an upper bound approach. The
failure mechanism considered (Figure 1) is based on
the classical Prandtl failure mechanism. Their original
contribution to this problem is to consider the dependence of the short term shear resistance of soft clays on
the local orientation of the failure surface. They propose an analytical upper bound solution based on this
3
1
2 + 2 (rock)
2
2
3
4
77
B1
D
i
Figure 1. Definition of geometric parameters.

1.4
1.2
Safety Factor
Figure 3. Comparison of free head pile group.
D=6m
D=8m
D=10m
D=12m
Hashash and Whittle (1996) using an advanced effective stress soil model, MIT-E3. The authors analysed
also a case of failure in Shanghai where the standard codes led to safety factors of more than 1.4 and
where this approach leads to a safety factor of 0.97,
explaining the basal instability.
Zhang and co-authors present a method for estimating the response of piles to lateral soil movements
induced by a nearby excavation.
For a single pile, the method is based on the classical
two-stage approach (Poulos & Chen 1997):
0.8
0.6
0. 9
0. 8
0.7
0.6
0.5
0.4
0. 3
Anisotropy Ratio
in a first step, the free-field soil movement must

be determined either by measurement or by calculation;
in a second step, these soil movements are imposed
to the piles through a Winkler subgrade reaction
model: the pile is represented by an elastic beam,
the pile-soil interaction is modeled using linear
elastic soil springs, the effect of axial load on
the pile is ignored. The Winkler subgrade reaction
equation is solved by a FD approach, permitting to
take into account heterogeneous soils.
Figure 2. Influence of D/H on the factor of safety

( = 18 kN/m3 , undrained shear strength Suv (z) = 0.33v ,
width of the excavation B = 15 m, depth of the excavation
H = 12 m).
kinematical mechanism and on the equation proposed

by Casagrande and Carillo (1944) for describing the
anisotropy of shear strength:
This classical method has been extended by the

authors to pile groups. In the case of pile groups, the
shielding effect of piles is modelled by superposing
to the free field soil movement the reduction of the
displacement due to neighbouring piles. This shielding effect is calculated using an attenuation function
based on simplified Mindlins equation.
The authors present a comparison of their approach
with centrifuge model tests and finite element simulation published by Leung et al (2000).
A comparison of calculated and measured bending
moments is given on Figures 3 and 4 in the case of
where Suh and Suv are obtained by undrained triaxial

compression and extension tests.
Figure 2 presents the results of a parametric study
on the evolution of the safety factor with the anisotropy
ratio, for different values of the embedment depth.
It appears clearly that the anisotropy ratio has more
influence than the embedment depth which increases
only slightly the safety factor. The authors studied also
the influence of the depth of the bedrock, when the
bedrock limits the extension of the failure mechanism.
This approach compared well with a 2D FE analysis
of adeep excavation in Boston Blue Clay a presented by
78
Figure 6. Forces acting on the TBM.
strut system, the load distribution between the different struts at the same level being calculated by
back-analysis.
As there are no direct load measurements, the magnitude of the loads obtained by this approach depends
strongly on the a priori hypothesis concerning the distribution of soil pressure acting on the wall. Therefore
this method can only give an indication on the relative
distribution of the loads.
Figure 4. Comparison of front pile in capped head pile

group.
3 TUNNELING
3.1 T.B.M. simulation
Chen and co-authors present an interesting paper on
the behaviour of aTBM when following a curved alignment. They propose a comprehensive numerical model
of an articulated shield which is an extension of a
kinematic shield model proposed for a single circular shield (Sugimoto & Sramoon 2002). Their model
is focused on the tunnel boring machine, considering
all the forces acting on the shield, such as for example
(Figure 6)
Figure 5. Example of strut system.
pile groups: this simplified model fits quite well with

the experimental results for free headed piles but there
are some differences in the case of capped piles. The
authors explain that this difference could be reduced
by using a non-linear elastic spring hypothesis, but
perhaps is it due to the Winklers hypothesis itself.
Finally, it must be noted that this method is not specific to excavations, and could be used for other works
inducing lateral soil movements, such as tunnelling or
embankments on soft clays.
Shi and co authors present a method for estimating
by back-analysis the strut loads in a complex concrete
strut system, such as the system presented on Figure 5.
The proposed method is based on measured displacements of the wall in the horizontal plane of the
the different jack thrust forces,

the pressure acting on the face,
the forces acting on the shield periphery.
These latter forces represent the interaction between
the shield and the surrounding soil. They are simulated
by a spring model. Therefore, in this model, the tunnelling operation is seen mainly from the point of view
of the TBM.
The simulation of the TBM behaviour is obtained by
imposing to the model the main operation parameters
of the shield.
A comparison between an observed and simulated
behaviour is presented in Figure 7. The actual shield
trajectory, in the vertical and horizontal plane, and in
79
x (m)
29.6
Vertical plane
29.7
29.8
end
start
29.9
20
40
60
80
100
Distance (m)
-6940
Horizontal plane
end
y (m)
-6920
R =20 m
-6900
-6880
-8640
Observed
Simulated
start
-8660
-8680
-8700
-8720
-8740
z (m)
Figure 8. Effect of radial seepage on the ground reaction

curve.
Figure 7. Simulated and observed behaviour.
extreme conditions of a sharp curve, is well simulated

by this very comprehensive model.
This model gives also the field of soil pressure acting on the shield, derived from the spring model, as
shown on this figure. In a further step, it could be interesting to use these calculated contact stresses between
the shield and the soil in a continuum model of the soil
mass for modelling the soil deformation and to check
if these calculated contact stresses lead to a realistic
simulation of observed displacements and settlement
troughs.
a semi impervious lining, having the same mechanical properties than the surrounding soil. A comparison
between the dry soil case and the fully drained case is
presented in Figure 8:
3.2
Despite the diverse simplifications of such an analytical model, this paper gives interesting indications
on the influence of seepage forces and shows clearly
that they should be taken into account for modelling the ground reaction curves and for assessing
the stability of the excavation.
In a second paper on ground reaction curves, Sozio
presents a 2D or 3D analytical model representing the
tunnel and the soil cover by a thick sphere or a thick
cylinder (Figure 9). The soil model is the classical
linear elastic-perfectly plastic Mohr-Coulomb model.
The originality of this model is that the gravity
forces are emulated by radial body forces. This enables
to take into account a limited cover depth, with a surface load and an internal pressure. The author proposes
to use this 3D model for a preliminary assessment
of the stability of the unlined length of a tunnel, the
problem being to estimate the radius of the sphere
equivalent to the tunnel unsupported heading.
Such analytical models are generally based on
restrictive hypotheses. It is the case for this model,
but it has to be highlighted that in his paper, the author
indicates very clearly the limitations of the proposed
models. It should be interesting to compare this model
to the classical approach based on the assumption of a
tunnel bored in an infinite soil mass.
in dry conditions, the pressure decreases rapidly

with the convergence, as the soil resistance is
mobilised,
when fully drained seepage is considered, there
is a marked increase of the convergence for a
given internal pressure, as the seepage forces do
not depend on the soil convergence and remain
constant.
Ground reaction curve
Shin and co-authors propose an analytical model of

the ground reaction curve taking into account the
seepage forces. The following classical hypotheses are
adopted:
The tunnel is bored in an infinite soil mass subjected
to a hydrostatic in situ stress,
The soil is linear elastic perfectly plastic with the
Mohr-Coulomb yield criteria,
Radial seepage forces are taken into account, as
indicated in the equilibrium equation:
The hydraulic gradients are calculated separately,

considering a steady state of seepage.
Based on these hypotheses, the authors propose an
analytical solution of the elasto-plastic state of stress,
expressed in terms of stress state and displacement.
Due to these assumptions, this model is more adapted
to deep tunnels.
The authors present an application of their model
for a 50 m deep tunnel, with a diameter of 5 meters.
Different cases are examined: fully drained or with
80
x
l
Djack
Dbr
lq
q
qw
Mjack
li
lu
Figure 10. Model used by Hoefsloot to represent the stage

construction of the tunnel lining.
Figure 9. Representation of the soil tunnel interaction.
The papers presented by Zhang & Wang and Lu et al

concern more specifically deep rock tunnels.
Zhang & Wang study the ground reaction in the case
of a pressure tunnel, the rock mass being not unloaded
but loaded by the internal pressure. In this specific
situation, quite far from urban tunnels in soft ground,
the softening of the rock considered by the authors can
lead to a broken zone around the tunnel.
Lu et al studied by 3D numerical simulations the
stability of different types of intersections between
deep mine tunnels and the influence of the construction
sequences. The method of simulation is not precisely
described. If this study is not directly applicable to
shallow tunnels in soft ground, some of their results
can be considered from a qualitative point of view
such as the fact that excavating towards the intersection appears more dangerous than excavating from the
excavation.
Figure 11. Groene hardt tunnel evolution of the bending

moment with the advance of the TBM.
the advancement of the TBM and becomes permanent

after about 60 meters. Despite the simple hypothesis of
a spring model, the evolution of the bending moment is
quite well modelled. Nevertheless this result has been
obtained by adjusting some parameters which are difficult to assess, such as the lining bending stiffness
and the effect of grouting (Talmon et al. 2008).
Finally, this analytical model, validated on field
measurements, shows that the staged construction of
the segmental lining in a straight alignment results in a
permanent longitudinal bending moment, that should
be considered in the design of the lining and for the
installation of the segments.
A second paper on the longitudinal behaviour of
segmented lining is presented by Zhu et al. The
authors examine the problem of the actual longitudinal stiffness of segmented linings which is one of the
parameters which was necessary to adjust in the model
proposed by Hoefsloot.
The assessment of the lining stiffness is based on
a 3D numerical model, composed of shell elements
(Figure 12) and joints with shear and normal stiffness
at all the interfaces between the individual elements.
The complete numerical model, loaded as a cantilever
beam, is compared to a simplified equivalent continuous model, which is not precisely described in the
paper.
The stiffness deduced from the numerical model
appears on their example to depend on the segment
length and to be lower than the stiffness obtained
3.3 Longitudinal behaviour of segmented lining

The paper proposed by Hoefsloot is based on the
observation that the staged construction of segmented
tunnel linings induces a permanent longitudinal bending moment in the lining. Based on solutions proposed
by Bogaards & Bakker (1999), and Bakker (2000), the
author proposes an analytical solution by considering
the segmented tunnel lining as a beam on an elastic
foundation.
The longitudinal loading scheme (Bending moment
and shear force from jack forces, shear force from steel
brushes, weight of lining segments, uniformly distributed load of limited length: back up train) advances
with the progress of the TBM (Figure 10). This analytical model has been built in a spreadsheet, and verified
using PLAXIS 2D.
The result of this model is compared with strain
measurements made in the lining of the GROENE
HART tunnel in the Netherlands. As illustrated on
Figure 11, a bending moment is induced in the lining by
81
Distance to the center of tunnel (m)
surface settlement(mm)
-30
0
Figure 12. Lining model.
-20
-10
10
20
30
20
40
60
80
100
2.U0
120
Peck's method
Stochastic medium theory
Verruijt and Booker's method (v=0.1)
Verruijt and Booker's method (v=0.2)
Figure 14. Settlement trough calculated using 3 different

methods (Zu and Liu).
U0/2
U0/2
The example presented on Figure 14 exhibits very large

differences, but it is in fact an extreme case, with a
cover depth of less than one meter for a tunnel 6 meters
in diameter.
They present also some observed settlement troughs
from Shanghai metro line 7 construction. It would have
been interesting to have more details on the tunnel
works leading to these settlement troughs and to compare the observed settlement troughs to the calculation
methods presented in the first part of the paper.
Figure 13. Convergence scheme proposed by Heli Bao et al.
by the equivalent continuous model. But both models are based on hypothesises concerning the joint
behaviour, which need to be measured or assessed
based on the observation of the actual behaviour of
full scale segmented linings.
3.4
3.5 Effect of vibrations

In their 2 complementary papers, Cui and co-authors
investigate experimentally the dynamic loading and
the development of pore pressure of saturated silty
clays near Shanghai subway Line No.2, during the passage of metro trains. On the observed site, settlements
exceeding 20 cm where observed, but no details on the
evolution of these settlements is given.
The experimental study is based on field observations and dynamic triaxial tests.
Based on in situ measurements, the authors propose an experimental law of attenuation of the dynamic
loading with the distance to the tunnel. This law, due to
the polynomial approximation adopted, has certainly
a limited domain of validity, and cannot be extrapolated outside the range of distance corresponding to
the measurement points.
They studied also the development of pore pressure with the dynamic loading, both in the field and
by triaxial tests; but the relation of field measurements results with dynamic triaxial tests is not clearly
described.
It would have been interesting to have more
details concerning the long term evolution of pore
pressure in situ, combined also with the evolution of
soil deformations during cycling loading tests, in order
to explain the observed large settlements.
Baimakhan and co-authors propose a coupled
approach (analytical and numerical) to determine the
Settlement troughs
There is generally a large consensus about the use of

the Gaussian type curve for describing the settlement
trough.
The direct estimation through elastic calculations or
by numerical simulations often leads to larger settlement troughs than observed. The result depends in fact
on different assumptions, one being the convergence
profile of the ground around the tunnel.
Heli Bao and his co-authors present an analytical solution, using conformal mapping of an elastic
half space. In order to fit with the observed settlement troughs, they propose an elliptical convergence
shape, based on the solution proposed by Park (1974),
as indicated on Figure 13.
This approach is compared with the observed settlements during the construction of a 6.2 m diameter
tunnel in Shanghai. The calculated settlement trough
fits quite well with the observed one. But no indication
is given concerning the assessment of the magnitude
of the convergence, the authors indicating simply the
gap between the TBM and the lining, gap which is in
fact certainly partially filled by grouting.
In their paper, Zu and Liu compared different methods for settlement trough assessment: Pecks empirical
approach (Peck 1969), stochastic medium theory, and
the solution proposed by Verruijt and Booker (1996).
82
stresses induced by earthquakes on the lining of

tunnels of subway lines.
Using a concept of homogeneous anisotropic elastic
medium they consider the effect of the succession of
different soil or rock layers.
They analyse in a more specific way the case of
tunnel or galleries with a longitudinal axis making an
angle with the major direction of anisotropy.
GENERAL PAPERS ON DESIGN

METHODS AND TOOLS
Figure 15. Distribution of estimated value.
characteristics of this statistical distribution as illustrated on Figure 15.

Such a distribution is important information for
tunnelling projects, based on a limited number of
investigation points, in order to evaluate the limits of
the design and for risk assessment.
One of the difficulties for multiplying 2D or 3D
numerical simulations on tunnelling projects is that
the pre-processing tasks are much time consuming.
Li and co-authors propose in their paper a methodology to derive a FEM model from numerical geological models, which are more and more used in the frame
of large geotechnical projects.
The figures presented Figure 16 show some of
the stages, beginning from the geological model and
ending to a 3D finite element mesh and where the
soil characteristics are imported from the geological
model.
In such an approach, coupling the geological model
with the geostatistical method presented in the previous paper could be certainly a very powerful tool for
helping the designer to test different hypothesis of the
soil parameters.
Negro presents the results of a comprehensive survey of current design practice in Brazil. The analysis
of this survey is based on the answers of 20 experts.
The topics of the survey concern the main aspects of
tunnel design:
Koungelis & Augarde compare, on an academic example, the results given by 2 different FE codes, Strand 7
and Plaxis. They investigate the effect of surface loading on wished in place tunnels in soft ground assuming
plane strain conditions.
The initial conditions and soil characteristics are
the same in all their simulations, except for dilatancy.
The two meshes used in their comparative study are
quite different: in Plaxis the mesh is more refined and
consists of fifteen noded triangular elements, and in
Strand 7 the mesh is coarser, and moreover consists of
simple 6 noded triangular elements.
The authors compare the changes in horizontal and
vertical diameters for different position of the surface
load. There are actually small differences although
there is a great difference between the refinement of
the meshes and the type of triangular elements.
Although no indication is given in the paper, one
can suppose that in this example, plastic zones are
certainly very limited or absent around the tunnel.
Therefore, this paper compares essentially the influence of the mesh refinement and of the type of element
in a linear elastic case, which explains the fact that only
minor difference is noted. Such comparisons should be
extended to less academic situations, where the tunnel construction is simulated and where small strain
behaviour is modelled or where large plastic zones are
mobilised.
Jeon and coauthors present an interesting communication on the use of geostatistical methods for
assessing the spatial distribution of the rock mass characteristic named RMR. They compare a method named
SIS (Juang et al., 2003, Feng et al., 2006) to the more
classical kriging (Marinoni, 2003; Pardo-Igzquiza
and Dowd, 2005).
The problem considered here, is how to assess, on
the base of limited bore holes, the ground characteristics along the tunnel alignment. The application
presented concerns deep rock tunnels, where geostatistical estimations of RMR around a tunnel are
compared.
Compared to kriging which gives a deterministic
value at each point considered, SIS gives a statistical distribution of the unknown value with different
Tunnel heading stability,

Settlement,
Damage to existing structures,
Lining design,
Account of ground water loading,
2D/3D FEM or FDM models,
Soil models in FEM/FDM models,
Soil investigations,
Monitoring.
The results of this survey are clearly summarized

and analysed in the paper by Negro, therefore, in this
report are given only some typical examples.
From this survey, it results that the typical scenario
for tunnel projects in Brazil is the following: tunnels
with equivalent diameter larger than 6 m, driven under
83
A large majority still use the linear elastic/plastic

Mohr Coulomb model which is well known, but certainly often not adapted to shallow urban tunnels where
the limitation of soil movements lead to small strains.
Very few or no plasticity will be mobilised, and the
model will be equivalent to a simple linear elastic
model which very poorly describes the small strain
behaviour of the soil. It can be added that this simple
model can be unsafe in coupled analysis, as the volume
changes in the soil are not correctly described.
This example shows that, with the large availability
of 2D or 3D Finite element or finite difference powerful codes, there is certainly a need for clarifications
concerning the types of soil models to be used in different situations and also a need for dissemination of
this knowledge.
The rich conclusions of such a survey could be certainly a good base for developing and enhancing good
practices in the field of tunnel design. For this reason
TC28 has proposed to launch national surveys based
on the example proposed by Negro, to be collected and
analysed for the next TC28 symposium.
5
Figure 16. Some stages from the geological model to the
3D FE model.
CONCLUSION
After excluding the 4 general papers, the 15 remaining papers allocated to this session, were related to 9
different specific topics.
This highlights that tunnels and deep excavations
are complex works, with strong interaction with their
environment, and that there is obviously a demand
for simple calculation tools addressing specific problems, easy to use, especially at the preliminary design
phases.
Concerning the calculation methods, it can be
noticed that 9 papers concern analytical or mixed
approaches as only 5 concern numerical methods.
Therefore, considering the limits of the analytical
approaches, due generally to the restrictive hypotheses
necessary to obtain a closed form solution, it should
be certainly useful now to develop simple numerical
tools, easy to use and time saving, dedicated to limited specific problems. These tools could be based on
existing codes and therefore able to take into account
advanced soil models and realistic geometries. Such
tools, after a comprehensive evaluation of their limits,
could be certainly useful for practitioners.
Another way to be considered is the development
of easy to use pre-processing tools, such as the example presented in this session, to facilitate the use of
complex 3D models.
And finally, it should be stressed that all these calculation methods have to be validated carefully and
in a scientific way against comprehensive measurements, and that the limitations of these models should
be clearly indicated.
Figure 17. Soil models used in numerical analyses.
mixed face condition, in cohesive soils, below water

table, and using sprayed concrete as lining (NATM).
Concerning the assessment of tunnel face stability, the survey shows that there is a large range of
methods currently used, and it is noticed that some of
these methods such as limit equilibrium, upper bound
solutions or empirical methods can be unsafe.
It is highlighted also that Practitioners are in fact
unhappy with the available methods for stability analysis, which certainly explains this broad range of
methods used in practice.
Another interesting result concerns the constitutive
models used in numerical analysis (Figure 17).
84
LIST OF PAPERS WITHIN SESSION
REFERENCES
Baimakhan, R.B., Danaev, N.T., Baimakhan, A.R.,

Salgaraeva, G.I., Rysbaeva, G.P., Kulmaganbetova, Zh.K.,
Avdarsolkyzy, S., Makhanova, A.A. & Dashdorj, S. Calculation of the three dimensional seismic stressed state
of Metro StationEscalatorOpen Line Tunnels system,
which is located in inclined stratified soft ground.
Bao, H., Zhang, D. & Huang, H. A Complex Variable Solution
for Tunneling-Induced Ground Movements in Clays.
Chen, J., Matsumoto, A. & Sugimoto, M. Simulation of articulated shield behavior at sharp curve by kinematic shield
model.
Cui, Z.D., Tang, Y.Q. & Zhang, X. Deformation and pore
pressure model of the saturated silty clay around a subway
tunnel.
Hoefsloot, F.J.M. Analytical solution of longitudinal
behaviour of tunnel lining.
Jeon, S., Hong, C. & You, K. Design of tunnel supporting
system using geostatistical methods.
Koungelis, D.K. & Augarde, C.E. Comparative study of
software tools on the effects of surface loads on tunnels.
Li, X.X., Zhu, H.H. & Lin, Y.L. Geologic model transforming method (GMTM) for numerical analysis modeling in
geotechnical engineering.
Lu, T.K., Guo, B.H., Cheng, L.C. & Wang, J. Review and
interpretation of intersection stability in deep underground based on numerical analysis.
Lu, Z.P. & Liu, G.B. Analysis of surface settlement due to the
construction of a shield tunnel in soft clay in Shanghai.
Negro, A. Urban Tunnels in Soil: Review of Current Design
Practice in Brazil.
Shi, Z., Bao, W., Li, J., Guo, W. & Zhu, J. A study on loads
from complex support system using simple 2D models.
Shin, Y.J., Shin, J.H. & Lee, I.M. Ground Reaction due to
Tunnelling below Groundwater Table.
Song, X.Y. & Huang, M.S. Basal Stability of Braced Excavations in K0-consolidated Soft Clay by Upper Bound
Method.
Sozio, L.E. Analytical Two and Three Dimension Models to Assess Stability and Deformation Magnitude of
Underground Excavations in Soil.
Tang, Y.Q., Cui, Z.D. & Zhang, X. Dynamic Response of Saturated Silty Clay around a Tunnel under Subway Vibration
Loading in Shanghai.
Zhang, C.R., Huang, M.S. & Liang, F.Y. Lateral Responses
of Piles due to Excavation-Induced Soil Movements.
Zhang, L.M. & Wang, Z.Q. Elastic-plastic analysis for surrounding rock of pressure tunnel with lining based on
material nonlinear softening.
Zhu, W., Kou, X., Zhong, X. & Huang, Z. Modification of Key
Parameters of Longitudinal Equivalent Model for Shield
Tunnel.
Bakker, K.J. 2000. Soil Retaining Structures. Rotterdam:

Balkema.
Bogaards, P.J. & Bakker, K.J. 1999. Longitudinal bending
in Geomechanics Proc. NUMOG VII: p. 317321.
Casarande, A. & Carillo, N. 1944. Shear failure of anisotropic
soil. J. of the Boston Society of Civil Engineers, 31(4).
Feng, Y., Tang, S. & Li, Z. 2006. Application of improved
sequential indicator simulation to spatial distribution
of forest type. Forest Ecology and Management 222:
391398.
prediction for deep excavations in soft clay. Journal of
Geotechnical and Geoenvironmental Engineering, ASCE,
122(6): 474486.
Juang, K., Chen, Y. & Lee, D. 2003. Using sequential indicator simulation to assess the uncertainty of delineating
heavy-metal contaminated soils. Environmental Pollution
127: 229238.
Leung, C.F., Chow, Y.K. & Shen, R.F. 2000. Behavior of
pile subjected to excavation-induced soil movement. Journal of Geotechnical and Geoenvironmental Engineering,
126(11): 947954.
Marinoni, O. 2003. Improving geological models using a
combined ordinaryindicator kriging approach. Engineering Geology 69(12): 3745.
Pardo-Igzquiza, E. & Dowd, P.A. 2005. Multiple indicator
cokriging with application to optimal sampling for environmental monitoring. Computers & Geosciences 31(1):
113.
Park, K.H. 2004. Elastic solution for tunneling-induced
ground movements in clays. International Journal of
Geomechanics 4(4): 310318.
Peck, R.B. 1969. Deep excavations and tunneling in soft
ground. Proceeding of 7th international conference on
soil mechanics and foundation engineering. Mexico City:
State of the Art Report.
Poulos, H.G. & Chen, L.T. 1997. Pile response due to
excavation-induced lateral soil movement. Journal of
Geotechnical and geoenvironmental engineering, ASCE,
123(2): 9499.
Sugimoto, M. & Sramoon, A. 2002. Theoretical model of
shield behavior during excavation I: Theory. Journal of
138155.
Talmon, A.M., Bezuijen, A. & Hoefsloot, F.J.M. 2008. Longitudinal tube bending due to grout pressures. Shanghai:
TC28.
Verruijt, A. & Booker, J.R. 1996. Surface settlements due
to deformation of a tunnel in an elastic half plane.
85
Analysis and numerical modeling of deep excavations

R.J. Finno
Department of Civil and Environmental Engineering, Northwestern University, Evanston, IL, USA
ABSTRACT: The sixteen papers comprising the general theme Analysis and Numerical Modeling of Deep
Excavations of the 6th International Symposium on Geotechnical Aspects of Underground Construction in Soft
Ground are summarized herein. General characteristics of all papers are presented as are brief summaries of each
paper. Most papers included presentation of results of finite element simulations and attendant comparisons with
various aspects of observed field performance. Some of the pitfalls for making comparisons between numerical
results and field observations of deep excavation performance are discussed briefly. In particular, the effects of
modeling construction details and selection of appropriate constitutive models are presented. Recommendations
are tendered regarding the essential information that should be conveyed in papers that present results of numerical
calculations.
INTRODUCTION
Table 1. Topics and number of papers.
Sixteen papers concerning analysis and numerical

modeling of deep excavations were published in the
Proceedings of the 6th International Symposium on
Geotechnical Aspects of Underground Construction in
Soft Ground. Of these papers, ten were presented orally
at the symposium.An overview of the papers is made to
provide a snapshot of the state of the practice as regards
to this topic. All papers are summarized and trends
in the contents are discussed. Because most papers
included presentation of results of finite element simulations and attendant comparisons with various aspects
of observed field performance of deep excavations,
general comments regarding factors that can be important in order to make an accurate prediction. Finally,
recommendations are tendered regarding the essential
information that should be conveyed in papers that
present results of numerical calculations.
Topic
No. of papers
Numerical analysis and measurements

Numerical analysis
Back-analysis
Measurements
Design
Stress path
Earth pressure
7
3
2
1
1
1
1
Table 2.
Summary of finite element codes used in papers.
Analyses
Number and code used
Three-dimensional simulations
2 -Plaxis 3D Foundations
2- FLAC3
2-Research code (?)
6-Plaxis
1-GeoTunnel
1-Research code (?)
Plane strain simulations
OVERVIEW
The sixteen papers covered the broad topics summarized in Table 1. The classification is somewhat
arbitrary and several papers could have fit into more
than one topic. It is clear, however, that the majority of
papers explicitly included comparisons of computed
results and some type of performance data.
Of these 16 papers, results of finite element analyses were reported in ten of them. Table 2 summarizes
the FE codes that were used. These results agree with
the authors experience that the commercial codes
FLAC and PLAXIS are most commonly used in both

geotechnical practice and research. The ease of use of
these codes has progressed to the point where threedimensional analyses have become more common, as
suggested by the number of such analyses presented
in the papers in this session.
While the use of finite element analyses has become
more common in practice, likely as a result of the
87
Table 3.
timely fashion in the lateral direction, thereby leaving

excessively large amounts of wall without lateral support. They compared the computed deformations of
the retaining structure, the horizontal displacement at
the top of SMW and the axial forces of steel pipe supports with the field observation data for both instances.
The authors concluded good agreement was shown
between the computed and observed results. From a
practical point of view, they showed that the normal
construction sequencing in case 1 resulted in a stable
and safe excavation; the axial forces were lower than
the alarm values and the displacement due to excavation were within the permissible range. In case 2,
however, the computed results showed the excavation
was close to becoming unstable, and measures had to
be taken to protect the retaining structure from failure.
Popa et al. presented Numerical modeling and
experimental measurements for a retaining wall of a
deep excavation in Bucharest, Romania. They summarized a case history of a diaphragm wall for a
deep basement of a new building in the center of
Bucharest. The excavation impacted a number of historic structures, leading to the use of top-down
techniques to support the excavation. The numerical
results obtained by plane strain FE simulation were
compared with measurements recorded during construction. Soil behavior was assumed to be that of a
Mohr-Coulomb material. The computed lateral displacements were 15% and 75% of the observed values,
depending on if the comparisons were made in an
area with or without a grouted wall not explicitly modeled in the FE simulations adjacent to the
diaphragm wall.
Schweiger et al. presented 3D finite element analysis of a deep excavation and comparison with in situ
measurements. The paper describes the results of FE
analyses using Plaxis of a deep excavation project in
clayey silt in Salzburg. The excavation was supported
by a diaphragm wall, a jet grout panel and three levels
of struts. The soil responses were represented by the
Hardening-Soil model. Because of insufficient information available at the time of design on the material
properties of the jet grout panel, the authors varied
its stiffness in a parametric study. The effect of taking
into account the stiffness of a cracked diaphragm wall
on the deformations also was investigated. In some
of the 3D calculations, the authors simulated nonperfect contact between the diaphragm wall and a strut
by means of a non-linear behaviour of the strut. The
evaluation of the results and comparison with in situ
measurements showed that analyses which took into
account the reduced stiffness of the diaphragm wall
due to cracking achieved the best agreement with the
measurements. Furthermore settlements of buildings
could be best reproduced by the three-dimensional
model, although the predicted settlements were not in
good agreement with the observations.
Constitutive model summary.
Model
No. of applications
Mohr-Coulomb
Hardening Soil
Duncan-Chang
6
3
1
improved i/o of commercial codes, the accuracy of the

results depends the faithful representation of activities
that induce stress changes in the ground during excavation and on the constitutive responses assumed in
the analyses. Simple models are easy to use, but are
limited in the types of computed responses that will
agree with observations.
A key aspect to applying finite element analysis to
practical problems in geotechnics remains the selection of the soil model and its individual parameters.
Of the ten papers that presented results of FE analyses,
six assumed soil responded as a Mohr-Coulomb material, as shown in Table 3. This assumption limits the
predictive capabilities of a FE simulation of deep excavations or tunneling in that elastic response is assumed
until the soil fails. Presumably, this choice was made
because of lack of detailed laboratory or field characterizations of the soils considered by the authors in
their papers. At the cost of simplicity, the HardeningSoil and Duncan-Chang models are non-linear and
can account for different responses in loading and
unloading. But these models also are limited in their
predictive capabilities in that they do not account for
the incremental non-linearity and small strain stiffness
responses that all soils exhibit. In any case, the factors
leading to the selection of a soil model should be discussed in a paper to put the results in context. In the
same spirit, the parameters and a rationale for their
selection also should be included in a paper.
3
3.1
SUMMARY OF PAPERS
Numerical analysis and measurements
Li and Huang presented Construction monitoring

and numerical simulation of an excavation with SMW
retaining structure. A SMW retaining structure was
used to support two long excavations in Shanghai.
Bearing and deformation mechanisms of the SMW
were analyzed briefly and the structural analysis of
SMW was discussed. Based on the in-situ excavation
procedures, the authors simulated construction of the
wall numerically using the FE code FLAC3D. They
represented soil behavior with a Mohr-Coulomb model
and considered two cases. Case 1 was the typical construction situation at the site wherein the supports were
installed in a timely fashion. Case 2 considered the
situation wherein the supports were not installed in a
88
higher than those commonly obtained from conventional triaxial tests, can be used as guideline for future
excavations in Bangkok.
Ota et al. presented Consideration of design
method for braced excavation based on monitoring
results. They compared observed and design values
of wall deflections at several cut-and-cover excavations through soft and sensitive clay ground at
the Osaka Subway Line No.8. A beam-spring model
was employed in the braced design method which
accounted for the characteristics of the Osaka soft
ground. While there was good agreement between the
observed data and design values in past results, the
observed wall deflections in this study were larger
than that expected for construction sites wherein the
excavations encountered 10 to 20 m thick, soft and
sensitive clay layer. The authors discuss how they evaluated the horizontal coefficient of subgrade reaction
kh on the excavation side of soft clay layer. The authors
make new recommendation regarding selection of kh ,
and show that the calculated wall movements with the
revised values agree with the observations. These recommendations are applicable to the soft and sensitive
Osaka clays.
Zhang and Huang presented Monitoring and

modeling of riverside large deep excavation-induced
ground movements in clays. They discussed a deep
excavation located at the Shanghai international passenger center that was 800 m long and 100150 m wide
with the depth of 13 m. The south side of the deep
excavation was within 4.6 m of a parallel flood wall
of the Huangpu River. The north side of the excavation was 5 m from a historic building. Because of
the differences in the conditions on the two sides of
the excavations, Plaxis FE analyses were conducted
which explicitly included both sides of the excavation, rather than a centerline symmetric condition. Soil
responses were assumed to correspond to that of a
Mohr-coulomb material. Computed differences of lateral wall movements on each side differed by as much
as 50%, as was verified by field observations made
during construction.
Hsi et al. presented Three-dimensional finite element analysis of diaphragm walls for top-down construction. They discussed the Tugun Bypass Tunnel
in Gold Coast, Australia. The tunnel was constructed
using diaphragm walls with the top-down cut-andcover method to allow simultaneous construction of
an airport runway extension above the tunnel, and
excavation of the tunnel beneath. The tunnel was
built in deep deposits of saturated, alluvial and estuarine soils with the toes of the walls founded in
soil deposits. There was a potential risk for differential settlements between the diaphragm wall panels,
caused by the runway fill placed over the tunnel
roof during excavation. Three-dimensional numerical
modeling was undertaken with Plaxis 3D Foundation
to predict the differential settlements of the tunnel
arising from the variable subsurface profile, staged
excavation and dewatering, non-uniform loading and
soil-structure interaction. Soil was assumed to behave
as a Hardening-Soil material. Settlements measured
after construction were within the range of those
computed with the finite element simulations.
Phienwej presented Ground movements in station
excavations of Bangkok first MRT. The characteristics of the lateral movements of the diaphragm walls
at excavations for 18 stations of the first Bangkok
underground MRT line were evaluated. Three modes
of deflected shapes of the walls were observed at different excavation depths, namely a cantilever mode
and braced modes with a bulge in soft clay and a bulge
in stiff clay. The ratio of maximum lateral wall deflection as a function of excavation depth and the ratio of
ground surface settlement to excavation depth and the
normalized ground surface settlement varied with the
mode of wall deflection. Undrained undrainedYoungs
moduli for a Mohr-Coulomb constitutive response for
different soil layers were back-calculated from wall
movement data of three selected stations using the 2D Plaxis FE code. The modulus values, which were
3.2
Numerical analysis
Li andYang presented the paper Numerical evaluation

of dewatering effect on deep excavation in soft clay.
They described a FLAC3D analysis that modeled
top-down construction of a 33.7 m deep underground
transformer substation in the downtown area of Shanghai. There are both unconfined and confined aquifers
on the site of this project and drainage by desiccation
in the foundation pit was adopted. Assuming a MohrCoulomb soil response, the effective stress methods
of analysis incorporated excavation and dewatering of
the foundation pit as part of the simulation of construction activities. The computed wall deflections, basal
heaves and surface settlements based on analyses that
did not consider dewatering were compared to those
that did. Results of analyses that considered leakage
through the wall and leakage between the aquifers are
presented as well. The analysis shows that although the
computed differences in lateral wall movement and
basal heave were small, due to the low permeability
of the soil, dewatering increased the amount of computed surface settlements as a result of drawdown of
the water outside the walls of the excavation.
Li et al. contributed the paper Analysis of the factors influencing foundation pit deformations. They
presented results of FE computations based upon
3-D Biots consolidation theory, assuming the soil
responded as a nonlinear Duncan-Changs material.
The finite element equations explicilty considered
the coupling of groundwater seepage and soil skeleton deformation during excavation. They presented
89
They employed Plaxis v8 and assumed the soil acted

as a Hardening-Soil material in their computations.
The authors stated that results of this approach indicated that the semi-reverse method is an effective way
to improve rigidity of the exterior support, control the
deformation of excavation, and ensure safety of the
surrounding buildings and pipelines.
results that showed the individual effects of the influence of soil permeability, rigidity and levels of lateral
supports, rigidity of retaining wall and excavation
duration on ground surface settlement, wall horizontal
displacement and basal heave of an excavation.
Siemiska-Lewandowsk and Mitew-Czajewska
presented the paper The effect of deep excavation
on surrounding ground and nearby structures. They
described problems related with the construction of a
29 m deep excavation of Nowy Swiat Station (S11)
of 2nd metro line in Warsaw. A critical section of
the project consisted of 7 stations and 6 running tunnels 6 km length in total. Running tunnels will be
constructed using TBM while the stations are to be
constructed using cut and cover techniques. Deep
excavation will be made with diaphragm walls supported by several levels of slabs and struts. They
presented results of 2-D Plaxis FE analyses in terms
of ground surface settlements, displacements of surrounding foundations and lateral wall movements,
assuming the soil behaves as a Mohr-Coulomb material. Additionally, settlements of the surface were
calculated above the TBM (running tunnels). Resulting values of settlements in both cases were discussed,
and formed the basis of design predictions that will be
verified during construction.
3.4
Zhang et al. contributed the paper GPS height application and gross error detection in foundation pit
monitoring. The authors introduced a deformation
monitoring model that combined traditional survey
technology and GPS measurements. They illustrated
foundation pit deformation monitoring based on their
experience of deep foundation pit construction of an
underground tunnel in Lishui Road, Hangzhou city.
When analyzing GPS height conversion to improve the
reliability of the GPS datum, they employed Dixons
test in the GPS datum mark to determine potential
height anomalies. The authors concluded that this
approach is a convenient way to search and delete raw
data that includes gross errors.
3.5
3.3
Measurements
Design
Chang contributed the paper Optimization design of

composite soil-nailing in loess excavation. Excavations through loess have unique characteristics compared with the others due to its structural properties
and collapsibility. To evaluate the mechanisms of support and to develop reasonable methods to design
composite soil-nailing in loess excavation, the authors
used results of finite element analysis to design a soil
nail support system. Their optimization design methods are based on the results of finite element analysis
apparently assuming Mohr-Coulomb soil responses.
They conducted the simulations to determine the regularity of deformation and the safety factor, as functions
of selected design variables. The authors justified their
methods by reporting that the lateral deformations of
the example excavation were limited to 16 mm.
Back analysis
Zghondi et al. presented the paper, Multi-criteria

procedure for the back-analysis of multi-supported
retaining walls. They described a numerical backanalysis procedure for multi-supported deep excavations based on the optimization of several indicators,
taking in account the forces in the struts and the
differential pressures derived from the wall displacement. The evaluation of the procedure is based on
results of 1 g small scale laboratory experiments on
semi-flexible retaining walls embedded in a Schneebelli material. The proposed numerical procedure was
applied to an excavation with 2 levels of struts with low
stiffness. The optimized Hardening Soil Model parameters form the basis of calculations of response of 14
different tested configurations. The results are compared with the classical methods, SubGrade Reaction
Method, Finite Element analysis with Mohr Coulomb
model and with the back-analysis using Hardening
Soil Model parameters based on biaxial tests results.
Zhang et al. contributed the paper Study on deformation laws under the construction of semi-reverse
method. Taking a 24.1-m-deep foundation pit of
Shanghai Metro Line 1 which uses the semi-reverse
construction process of three open excavating-one
tunneling as an example, they determined deformation laws of a foundation pit under the construction of
a semi-reverse method based on analysis of field monitoring data and forward and back analyses methods.
3.6
Stress path
Zhou et al. contributed the paper Comparison of theory and test on excavation causing the variation of soil
mass strength. In view of the characteristic unloading caused by excavations, they deduced the strength
ratio of the unloaded soil to soil subjected to compression loadings. Laboratory tests simulating excavation
were carried out based on Hvorslevs strength theory.
By comparing theoretical results with the laboratory
data, they concluded that the soil mass is overconsolidated. As a consequence, the authors stated that the soil
microstructure is damaged, and the soil mass strength
90
impacts the formulation required. Clough and Mana

(1976) and field data have shown that for excavations through saturated clays with typical excavation
periods of several months, the clays remain essentially undrained with little dissipation of excess pore
pressures.
For undrained conditions, one can employ either a
coupled finite element formulation where both displacements and pore water pressures are solved for
explicitly (e.g. Carter et al. 1979) or a penalty formulation (e.g. Hughes 1980) wherein the bulk modulus of
water or a sufficiently large number that depends on
the precision of the machine making the computation is added to the diagonal terms in the element stiffness
matrix during global matrix assembly. This additional
term constrains the volumetric strain to nearly zero,
i.e., undrained. In both these approaches, the constitutive response of the soil is defined in terms of effective
stress parameters. A simpler, alternate approach is to
define undrained constitutive response in terms of total
stress parameters, with care being taken to make the
diagonal terms of the element stiffness matrix large,
typically by using a Poissons ratio close to 0.5. In this
case, a Youngs modulus corresponds to an undrained
value and failure is expressed in terms of an undrained
shear strength, Su (e.g., = 0 and c = Su ).
However, there may be cases (e.g., ORourke and
ODonnell 1997) where substantial delays during construction occur and excess pore pressures partially
dissipate, and in these cases one must use a mixed formulation to account for the pore water effects. When
using top-down techniques to excavate, it can take up to
several years to reach final grade for large excavations,
and hence partially drained conditions would apply
therein, requiring a coupled finite element simulation.
is reduced in the unloading process. The authors concluded that analyses of the results are helpful to the
understanding of the effect of excavation unloading
on the variation of the soil mass strength.
3.7
Earth pressure
Lin and Lee contributed the paper A simplified spatial methodology of earth pressure against retaining
piles of pile-row retaining structure. When using a
pile-row retaining structure to support excavation, the
authors stressed the importance of obtaining the magnitude and distribution of the earth pressure against
the retaining piles. Based on the mode of failure, a
new methodology is proposed to evaluate the earth
pressure against the retaining piles of such a structure. In the proposed method, both the spatial effect
and intermediate principal stress effect are considered.
The authors provide an example of the methodology
applied to engineering practice. They demonstrated
that the strength theory has more influence on earth
pressure.
4
COMMENTS REGARDING COMPUTED

AND OBSERVED RESULTS
Many factors affect ground movements caused by

excavations, including stratigraphy, soil properties,
support system details, construction activities, contractual arrangements and workmanship. In this theme,
most papers described numerical simulations that
analyzed ground response arising from excavation.
Finite element predictions always contain uncertainties related to soil properties, support system details
and construction procedures. Furthermore, while supported excavations commonly are simulated numerically by modeling cycles of excavation and support
installation, it generally is necessary to simulate all
aspects of the construction process that affect the stress
conditions around the cut to obtain an accurate prediction of behavior. This may involve simulating previous
construction activities at the site, installation of the
supporting wall and any deep foundation elements, as
well as the removal of cross-lot supports or detensioning of tiedback ground anchors. Issues of time effects
caused by hydrodynamic effects or material responses
may be important. The following sections summarize some of the factors that may impact computed
responses of ground movements associated with excavations. Proper consideration must be given to such
factors when making such analyses, as well as when
critically evaluating published results of the same.
4.1
4.2 Initial conditions

A reasonable prediction of the ground response to
construction of a deep excavation starts with a good
estimate of the initial stress conditions, in terms of
both effective stresses and pore water pressures. The
effective stress conditions for excavations in welldeveloped urban areas rarely correspond to at-rest
conditions because of the myriad past uses of the land.
Existence of deep foundations and/or basements from
abandoned buildings and nearby tunnels changes the
effective stresses from at-rest conditions prior to the
start of excavation. For example, Calvello and Finno
(2003) showed that an accurate computation of movements associated with an excavation could only be
achieved when all the pre-excavation activities affecting the site were modeled explicitly. They used the
case of the excavation for the Chicago-State subway
renovation project (Finno et al. 2002), wherein construction of both a tunnel and a school impacted the
ground stresses prior to the subway renovation project.
Drainage conditions
An important preliminary decision in any analysis is to

match the expected field drainage conditions, which
91
To approximate the effects of this arching when making such an analysis, an equivalent fluid pressure,
generally higher than the level of the fluid during construction, can be applied to the walls of the trench to
maintain stability. Thus, some degree of empiricism is
required to consider these installation effects in a plane
strain analysis. One can back-calculate an equivalent
fluid pressure corresponding to the observed ground
response if good records of lateral movements close
to the wall are recorded during construction. More
data of this type are needed before any recommendations can be made regarding magnitudes of appropriate
equivalent pressures.
The effects of installing a sheet pile wall are different than those of a diaphragm wall, yet the effects
on observed responses also can be significant. In
this case, ground movements may arise from transient vibrations developed as the sheeting is driven or
vibrated into place and from the physical displacement
of the ground by the sheeting. The former mechanism is of practical importance when installing the
sheeting through loose to medium dense sands, and
can be estimated by procedures proposed by Clough
et al. (1989). However, these effects are not included
in finite element simulations. The latter mechanism
in clays was illustrated by Finno et al. (1988). In this
case, the soil was displaced away from the sheeting as
it was installed. This movement was accompanied by
an increase in pore water pressure and a ground surface
heave. As the excess pore water pressures dissipated,
the ground settled. The maximum lateral movement
and surface heave was equal to one-half the equivalent
width of the sheet pile wall, defined as the crosssectional area of the sheet pile section per unit length of
wall. Sheet-pile installation can be simulated in plane
strain by using procedures summarized in Finno and
Tu (2006).
In addition to the movements that occur as a wall
is installed, installing the walls can have a large influence on subsequent movements, especially if the walls
are installed relatively close to each other, as may be
the case in a cut-and-cover excavation for a tunnel.
Sabatini (1991) conducted a parametric study as a
function of the depth, H, to width, B, of an excavation, wherein the effects of sheet-pile wall installation
in clays were compared with simulations where the
walls were wished into place. The results of the study
are shown in Figure 1 where the computed normalized
maximum lateral movements, H(max) /H, are plotted
versus H/B.
It is apparent for wide excavations (H/B 0.25)
that the decision to include installation effects in
a simulation is not critical. However, these effects
become pronounced for narrow excavations (H/B 1)
and should be explicitly considered. The results also
show that for the wished-in-place case when the
sheet-pile installation effects are ignored, the lateral
Ignoring these effects made a difference of a factor of

3 in the computed lateral movements.
One also must take care when defining the initial ground water conditions. Even in cases where
the ground water level is not affected by near surface
construction activities, non-hydrostatic conditions can
exist for a variety of reasons. For example, Finno et al.
(1989) presented pneumatic piezometer data that indicated the presence of a downward gradient within a
20 m thick sequence of saturated clays. This downward flow arose from a gradual lowering since the
1950s of the water level in the upper rock aquifer in
the Chicago area. A non-hydrostatic water condition
affects the magnitude of the effective stresses at the
start of an excavation project.
An engineer has two choices to define such conditions to measure the in situ conditions directly or
to simulate all the past construction activities at a site
starting from appropriate at-rest conditions. Because
both approaches present challenges, it is advantageous
to do both to provide some redundancy in the input. In
any case, careful evaluation of the initial conditions
is required when numerically simulating supported
excavation projects, especially in urban areas.
4.3 Wall installation
Many times the effects of installing a wall are ignored
in a finite element simulation and the wall is wishedinto-place with no change in the stress conditions
in the ground or any attendant ground movements.
However, there are abundant data that show ground
movements may develop as a wall is installed.
ORourke and Clough (1990) presented data that
summarized observed settlements that arose during
installation of five diaphragm walls. They noted settlements as large as 0.12% of the depth of the trench.
These effects can be evaluated by 3-dimensional modeling of the construction process (e.g., Gourvenec and
Powrie 1999), but not without several caveats. The
specific gravity of the supporting fluid usually varies
during excavation of a panel as a result of excavated
solids becoming suspended increasing the specific
gravity above the value of the water and bentonite mixture and subsequently decreasing when the slurry is
cleaned prior to the concrete being tremied into place.
Consequently, it is difficult to select one value that represents an average condition. Furthermore the effects
of the fluid concrete on the stresses in the surrounding
soil depend upon how quickly the concrete hardens relative to its placement rate. Some guidance in selecting
the fresh concrete pressure is provided by Lings et al.
(1994).
It is less straightforward when modeling diaphragm
wall installation effects in a plane strain analysis
because the arching caused by the excavation of individual panels cannot be taken directly into account.
92
H(MAX)/H [%]
support installation on one of the walls of the excavation for selected days after construction started. Also
shown are the locations of two inclinometers placed
several meters behind the wall. If one is making a
computation assuming plane strain conditions, then
it is clear that one must judiciously choose data sets
so that planar conditions would be applicable to those
selected.
Even when a sufficiently extensive horizontal excavated surface is identified, 3-dimensional effects may
still arise from the higher stiffness at the corners of an
excavation. These boundary conditions lead to smaller
ground movements near the corners and larger ground
movements towards the middle of the excavation wall.
Another, and less recognized, consequence of the corner stiffening effects is the maximum movement near
the center of an excavation wall may not correspond
to that found from a conventional plane strain simulation of the excavation, i.e., 3-D and plane strain
simulations of the excavation do not yield the same
movement at the center portion of the excavation, even
if the movements in the center are perpendicular to the
wall (Ou et al. 1996). This affect can be quantified by
the plane strain ratio, PSR, defined herein as the maximum movement in the center of an excavation wall
computed by 3-D analyses divided by that computed
by a plane strain simulation. Finno et al. (2007) developed the following expression for PSR from the results
of a finite element parametric study of excavations
through clay:
Sheet-pile effects
No sheet-pile effects
0.25
0.5
0.75
H/B
Figure 1. Effects of sheet-pile installation on computed
lateral movements.
0
North
South
Scale (m)
Strut 6
(79)
Strut 5
(74)
Strut 4
(60)
Strut 3
(59)
Strut 2
(58)
Strut 1
(56)
-5
Day 28
Initial
Grade
Day 57
Day 58
Day 60
Inclinometer 1
Location
15
East
Escalator Pit
Inclinometer 2
Location
Note:
Depth (m)
10
Day 75
Day 80
Day 88
Day 89
Day 95
Day 110
Day 113
20
Upper level tiebacks stressed between Day 81 and Day 105

Lower level tiebacks stressed between Day 98 and Day 113
Day 163
Tiebacks
Struts
25
40
35
30
25
20
15
10
-5
Distance From South End of Excavation (m)
Figure 2. Construction progress at excavation in Chicago

(Finno et al. 2002).
movements are larger for wider excavations, a similar

trend reported by Mana and Clough (1981).
Sheet-pile installation has two main effects: the soil
adjacent to the excavation is preloaded and the shear
strength on the passive side is (partially) mobilized
prior to the beginning of the cycles of excavation.
Wall installation tends to preload the soil on the active
side of the excavation as a result of the reduction in
shear stress at approximately constant mean normal
effective stress. This mechanism provides the soil outside the walls with more available shearing resistance
when the cycles of excavation start. However, the soil
between the walls has less available passive resistance
as a result of the preloading and this promotes the
larger movements during excavation as compared to
the case of ignoring the sheet-pile effects (Finno and
Nerby 1989).
4.4
where L is the excavation length along the side where

the movement occurs, B is the other areal dimension,
and He is the excavation depth. The value of C depends
on the factor of safety against basal heave, FSBH , and
is taken as:
The value of k depends on the support system stiffness

and is taken as:
where EI is the bending stiffness of the wall, is the

total unit weight of the soil and h is the average vertical
spacing between supports. When L/He is greater than
6, the PSR is equal to 1 and results of plane strain simulations yield the same displacements in the center of
an excavation as those computed by a 3-D simulation.
When L/He is less than 6, the displacement computed
from the results of a plane strain analysis will be larger
than that from a 3-D analysis. When conducting an
inverse analysis of an excavation with a plane strain
Plane strain versus 3-dimensional analyses
Figure 2 illustrates some of the challenges of using

field observations to calibrate numerical models of
any kind, even when detailed records exist. This figure
summarizes the construction progress at the ChicagoState excavation in terms of excavation surface and
93
simulation when L/He is relatively small, the effects

of this corner stiffening is that an optimized stiffness
parameter will be larger than it really is because of
the lack of the corner stiffening in the plane strain
analysis. This effect becomes greater as an excavation
is deepened because the L/He value increases as the
excavated grade is lowered. This trend was observed
in the optimized parameters for the deeper strata at the
Chicago-State subway renovation excavation (Finno
and Calvello 2005).
5
Secant Shear Modulus, Gsec (MPa)
90
GBE Range
70
60
unloading
50
loading
40
30
20
10
0
CONSTITUTIVE MODEL
CONSIDERATIONS
0.0001
0.001
0.01
0.1
Local Triaxial Shear Strain, sl (%)
Figure 3. Secant shear modulus as a function of direction

of loading.
When one undertakes a numerical simulation of a deep

supported excavation, one of the key decisions made
early in the process is the selection of the constitutive
model. In general, this selection should be compatible
with the objectives of the analysis. If the results form
the basis of a prediction that will be updated based on
field performance data, then the types of field data that
form the basis of the comparison will impact the applicability of a particular model. Possibilities include
lateral movements based on inclinometers, vertical
movements at various depths and distances from an
excavation wall, forces in structural support elements,
pore water pressures or any combinations of these data.
When used for a case where control of ground movements is a key design consideration, the constitutive
model must be able to reproduce the soil response at
appropriate strain levels to the imposed loadings.
5.1
FB3AL1
FB1TC2
FB1CMS2
FB1RTC2
FB1RTE1
FB1TE1
FB2CMSE1
80
axisymmetric conditions. Bender element (BE) tests

were conducted during all phases of the tests. The
secant shear moduli are plotted versus triaxial shear
strain in Figure 3 for natural specimens whose stress
probes involved changes in the shear stress q. The
overconsolidation ratio of these specimens was 1.7,
so if one assumes the response is isotropic and elastoplastic, then G should be the same for at least the initial
portion of all curves. The stress probes wherein q and
the stress ratio, = q/p , is increased ( loading)
are clearly softer than those where q and initially
decrease ( unloading). There are no obvious zones
of constant Gsec at shear strains greater than 0.002%,
and thus no elastic zone is observed in these data for
strain levels. Complete details and results of the testing
program are presented by Cho (2007).
Burland (1989) suggested that working strain levels
in soil around well-designed tunnels and foundations
are on the order of 0.1 %. If one uses data collected
with conventional triaxial equipments to discern the
soil responses, one can reliably measure strains 0.1%
or higher. Thus in many practical situations, it is not
possible to accurately incorporate site-specific small
strain non-linearity into a constitutive model based on
conventionally-derived laboratory data.
Incremental non-linearity
It is useful to recognize that soil is an incrementally

nonlinear material, i.e., its stiffness depends on loading direction and strain level. Real soils are neither
linear elastic nor elastic-plastic, but exhibit complex
behavior characterized by zones of high stiffness at
very small strains, followed by decreasing stiffness
with increasing strain. This behavior under static loading was identified through back-analysis of foundation
and excavation movements in the United Kingdom
(Burland, 1989). The recognition of zones of high
initial stiffness under typical field conditions was followed by efforts to measure this ubiquitous behavior
in the laboratory for various types of soil (e.g., Jardine
et al. 1984; Clayton and Heymann 2001; Santagata
et al. 2005; Calisto and Calebresi 1998, Cho 2007).
To illustrate this behavior, Figure 3 shows the
results of drained, triaxial stress probes conducted on
specimens cut from block samples of lightly overconsolidated glacial clays obtained at an excavation
in Evanston, IL. Each specimen was reconsolidated
under K0 conditions to the in-situ vertical effective

stress v0
, subjected to a 36 hour K0 creep cycle,
followed by directional stress probing under drained
5.2 Model selection

There are a number of models reported in literature wherein the variation of small strain nonlinearity
can be represented, e.g., a three-surface kinematic
model develop for stiff London clay (Stallebrass and
Taylor 1997), MIT-E3 (Whittle and Kavvadas 1994),
hypoplasticity models (e.g. Viggiani and Tamagnini
1999), and a directional stiffness model (Tu 2007). To
derive the necessary parameters, these models require
either detailed experimental results or experience with
the model in a given geology. While the models can
be implemented in material libraries in some commercial finite element codes, these routines are not readily
94
Maximum Lateral Movement

(mm)
120
100
80
60
40
20
0
0
0.2
0.4
0.6
0.8
1.2
Maximum Shear Strain (%)
Figure 5. Relation between maximum wall movement and

shear strain.
strain levels. In this case, the fact that small strain

non-linearity is not explicitly considered will not have
a large impact on the computed horizontal wall displacements because the maximum horizontal movement at the wall is dominated by the larger strains in
the soil mass. These computed movements would be
compatible with those measured by an inclinometer
located close to the wall.
However, if one needs to obtain an accurate representation of the distribution of ground movements with
distance from the wall, then this approach of selecting strain-level appropriate elastic parameters is not
applicable. Small strain non-linearity of soil must be
explicitly considered to find the extent of the settlement because the strains in the area of interest vary
from the maximum value to zero. As a consequence,
many cases reported in literature indicate computed
wall movements agree reasonably well with observed
values, but the results from the same computations do
not accurately reflect the distribution of settlements.
Indeed, this was the case in several papers presented as
part of this theme. Good agreement at distances away
from a wall can be obtained only if the small stain
non-linearity of the soil is adequately represented in
the constitutive model.
The relation between lateral wall displacements and
shear strain levels in the soil behind the wall can
be evaluated from results of displacement-controlled
finite element simulations. Similar to the results shown
in Figure 4, different displacement profiles were studied by imposing lateral wall displacements and settlement profiles, representing conditions with maximum
lateral movements at the excavated surface, cantilever
movements, deep-seated movements and combination
of the latter two (Andrianis 2006). The stratigraphies
used in the models were based on typical Chicago
soils. The results in Figure 5 show that the relationship between maximum shear strain behind the wall
and maximum displacement of the wall is almost linear
for lateral wall displacements between 10 and 110 mm.
Figure 5 also shows that the results form a narrow
Figure 4. Shear strains behind excavation: 57 mm maximum lateral movement (contours in %).
available to most practitioners. Thus for most current

practical applications, one uses simpler, elasto-plastic
models contained in material libraries in commercial
codes.
For these models, a key decision is to select the
elastic parameters that are representative of the
secant values that correspond to the predominant strain
levels in the soil mass. Examples of the strain levels
behind a wall for an excavation with a maximum lateral wall movement of 57 mm are shown in Figure 5.
These strain levels were computed based on the results
of displacement-controlled simulations where the lateral wall movements and surface settlements were
incrementally applied to the boundaries of a finite element mesh. The patterns of movements were typical
of excavations through clays, and were based on those
observed at an excavation made through Chicago clays
(Finno and Blackburn 2005). Because the simulations
were displacement-controlled, the computed strains do
not depend on the assumed constitutive behavior.
As can be seen in Figure 4, maximum shear strains
as high as 0.7% occur when 57 mm of maximum
wall movement develop. The maximum strains are
proportional maximum lateral wall movement; for
example, when 26 mm maximum lateral wall movement develops, the maximum shear strain is about
0.35%. These latter strain levels can be accurately
measured in conventional triaxial testing, and thus if
one can obtain specimens of sufficiently high quality,
then secant moduli corresponding to these strain levels
can be determined via conventional laboratory testing.
Because the maximum horizontal wall displacement
can be thought of as a summation of the horizontal
strains behind a wall, the maximum wall movements
can be accurately calculated with a selection of elastic parameters that corresponds to these expected
95
7 Construction records, simulation steps and details

of how each construction activity was idealized in
the finite element simulation.
band, suggesting that the relation between strain and

wall displacement is not greatly affected by the type
of movement.
Figure 5 can be used to estimate shear strains for a
specified maximum wall movement. With this value of
shear strain, the secant shear moduli for use in conventional elasto-plastic models can be estimated based on
strain-stress data from high quality laboratory experiments. The values of maximum shear strains, even in
the cases with the relatively low values of displacements, are about 0.2% and increase as the specified
displacement becomes larger. This is important when
one determines soil stiffness in the laboratory. Conventional soil testing without internal instrumentation
allows one to accurately measure strains as low as
0.1%. Thus for many cases, the secant shear moduli can be determined from conventional laboratory
tests on high quality samples. However, if strain levels
are 0.1% or less, then one must select these moduli
from test results based on internally-measured strains
in equipment not normally available in commercial
laboratories.
In summary, using a simulation based on conventional elasto-plastic models limits the type and
location of the data that can be used as observations in an inverse analysis. Both vertical and lateral
movements measured at some distance from a wall
cannot be calculated accurately in this case because
the variation of stiffness with strain levels must be adequately represented in the soil model. Only the lateral
movements close to a support wall can be reasonably
computed with conventional models since that result is
dominated by the zones of high strains behind the wall.
6
Finally, comparisons between computed and

observed results, as well as a discussion of the comparisons, should be included.
REFERENCES
Andrianis, G.E. 2006. Excavation-induced Strains and Cantilever Deflections in Compressible Clays. MS thesis,
Northwestern University, Evanston, IL.
Burland, J.B. 1989. Small is beautiful the stiffness of soils
at small strains: Ninth Laurits Bjerrum Memorial Lecture.
Canadian Geotechnical Journal 26: 499516.
Calisto, L. & Calebresi, G. 1998. Mechanical behavior of a
natural soft clay. Geotechnique 48 (4): 495513.
Calvello, M. & Finno, R.J. 2003. Modeling excavations in
urban areas: effects of past activities. Italian Geotechnical
Journal 37(4): 923.
Carter, J.P., Booker, J.R. & Small, J.C. 1979. The analysis of
finite elasto-plastic consolidation. International Journal
for Numerical and Analytical Methods in Geomechanics
3: 107129.
Cho, W.J. 2007. Recent stress history effects on compressible Chicago glacial clay. PhD thesis, Northwestern
University, Evanston, IL.
Clayton, C.R.I. & Heymann, G. 2001. Stiffness of geomaterials at very small strains. Geotechnique 51(3): 245255.
Clough, G.W. & Mana, A.I. 1976. Lessons learned in finite
element analysis of temporary excavations. Proceedings,
2nd International Conference on Numerical Methods in
Geomechanics, ASCE, Vol. I: 496510.
Clough, G.W., Smith, E.M. & Sweeney, B.P. 1989. Movement control of excavation support systems by iterative
design. Current Principles and Practices, Foundation
Engineering Congress, Vol. 2, ASCE: 869884.
Finno, R.J., Atmatzidis, D.K. & Nerby, S.M. 1988. Ground
response to sheet-pile installation in clay. Proceedings,
Second International Conference on Case Histories in
Geotechnical Engineering, St. Louis, MO.
Finno, R.J., Atmatzidis, D.K. & Perkins, S.B. 1989. Observed
Performance of a Deep Excavation in Clay. Journal of
Geotechnical Engineering, ASCE, 115 (8): 10451064.
Finno, R.J. & Blackburn, J.T. 2005. Automated monitoring of supported excavations. Proceedings, 13th Great
Lakes Geotechnical and Geoenvironmental Conference,
Geotechnical Applications for Transportation Infrastructure, GPP 3, ASCE, Milwaukee, WI.: 112.
Finno, R.J., Blackburn, J.T. & Roboski, J.F. 2007. Threedimensional Effects for Supported Excavations in Clay.
Journal of Geotechnical and Geoenvironmental Engineering, ASCE, 133(1): 3036.
Finno, R.J., Bryson, L.S. & Calvello, M. 2002. Performance
of a stiff support system in soft clay. Journal of Geotechnical and Geoenvironmental Engineering, ASCE, 128(8):
660671.
Finno, R.J. & Calvello, M. 2005. Supported excavations: the
observational method and inverse modeling. Journal of
Geotechnical and Geoenvironmental Engineering. ASCE,
131 (7).
CONCLUDING REMARKS
The papers presented at the symposium included

widely variable levels of information regarding the
details of the finite element analyses. As such, the
author tentatively proposes that the following information be included in any paper describing the results
of any finite element simulation of geotechnicallyrelated construction.
1 The finite element code used.
2 The assumed drainage conditions, e.g., drained,
undrained or partially drained.
3 The dimensionality of the problem, e.g., plain
strain, axisymmetric or three-dimensional.
4 The constitutive model(s) employed for both soils
and structural elements.
5 The parameters for each material and a discussion
of the basis of their selection.
6 A description of the mesh, including boundary conditions and type of elements used for soil, structural
components and interfaces.
96
Finno, R.J. & Nerby, S.M. 1989. Saturated Clay Response

During Braced Cut Construction. Journal of Geotechnical
Engineering, ASCE, 115(8): 10651085.
Finno, R.J. & Tu, X. 2006. Selected Topics in Numerical Simulation of Supported Excavations. Numerical Modeling
of Construction Processes in Geotechnical Engineering for Urban Environment, International Conference of
Construction Processes in Geotechnical Engineering for
Urban Environment, Th. Triantafyllidis, ed., Bochum,
Germany, Taylor & Francis, London: 320.
Gourvenec, S.M. & Powrie, W. 1999. Three-dimensional
finite element analysis of diaphragm wall installation.
Geotechnique, 49(6): 801823.
Hughes, T.J.R. 1980. Generalization of selective integration
procedures to anisotropic and nonlinear media. International Journal of Numerical Methods in Engineering 15
(9): 14131418.
Jardine, R.J., Symes, M.J. & Burland, J.B. 1984. The measurement of small strain stiffness in the triaxial apparatus.
Geotechnique 34 (3): 323340.
Lings, M.L., Ng, C.W.W. & Nash, D.F.T. 1994. The lateral
pressure of wet concrete in diaphragm wall panels cast
under bentonite. Geotechnical Engineering, Proceedings
of the Institution of Civil Engineers, 107: 163172.
Mana,A.I. & Cough, G.W. 1981. Prediction of movements for
braced cut in clay. Journal of Geotechnical Engineering,
ASCE, New York, 107(8): 759777.
ORourke, T.D. & Clough, G.W. 1990. Construction induced
movements of insitu walls. Proceedings, Design and Performance of Earth Retaining Structures, Lambe, P.C. and
Hansen L.A. (eds). ASCE: 439470.
ORourke, T.D. & ODonnell, C.J. 1997. Deep rotational

stability of tiedback excavations in clay. Journal of
Geotechnical Engineering, ASCE, 123(6): 506515.
Ou, C.Y., Chiou, D.C. & Wu, T.S. 1996. Three-dimensional
finite element analysis of deep excavations. Journal of
Geotechnical Engineering, ASCE, 122(5): 473483.
Sabatini, P.J. 1991. Sheet-pile installation effects on computed
ground response for braced excavations in soft to medium
clays. MS thesis, Northwestern University, Evanston, IL.
Santagata, M., Germaine, J.T. & Ladd, C.C. 2005. Factors
Affecting the Initial Stiffness of Cohesive Soils. Journal of
131(4): 430441.
Stallebrass, S.E. & Taylor, R.N. 1997. The development and
evaluation of a constitutive model for the prediction of
ground movements in overconsolidated clay. Geotechnique 47(2): 235253.
Tu, X. 2007. Tangent stiffness model for clays including small
strain non-linearity. PhD thesis, Northwestern University,
Evanston, IL.
Viggiani, G. & Tamagnini, C. 1999. Hypoplasticity for
modeling soil non-linearity in excavation problems. Prefailure Deformation Characteristics of Geomaterials,
M. Jamiolkowski, M, Lancellotta, R. and Lo Presti, D.
(eds.), Balkema, Rotterdam: 581588.
Whittle, A.J. & Kavvadas, M.J. 1994. Formulation of MIT-E3
constitutive model for overconsolidated clays. Journal of
120(1): 173198.
97
Construction method, ground treatment, and conditioning for tunneling

T. Hashimoto & B. Ye
Geo-Research Institute, Osaka, Japan
G.L. Ye
Department of Civil Engineering, Shanghai Jiaotong University, Shanghai, P.R. China
ABSTRACT: This general report reviews a selected group of papers of Session 2 which is related to
Construction Method, Ground Treatment, and Conditioning for Tunneling. The papers are divided into 5 groups
based on their topics: (1) construction methods with case studies, (2) ground treatment, (3) load and pressure,
(4) conditioning additives for EPB, (5) others. Before reviewing, the geotechnical aspects in these fields are firstly
summarized, and then the essences of these papers are presented. The deficiencies and future developments are
also discussed.
Table 1.
INTRODUCTION
Session 2 includes 19 papers from China, Japan, UK,

the Netherlands, Germany, Iran, Slovakia, Argentina
and Brazil. Especially, Tongji University, Shanghai
and GeoDelft, the Netherlands contribute to present
some papers respectively. It is because that Shanghai
and the Netherlands have been performed many tunnels during the past decade in the soft ground. These
papers are divided into 5 groups and 8 subgroups
based on their topics, which are shown in Table 1.
Although all these paper has contributed to some specified aspect of construction method, ground treatment,
and conditioning for tunneling, some papers with significant importance are selected to be reviewed in this
General Report. The review will be carried out according to the grouping of the paper. Before the review,
the geotechnical aspects in these fields are firstly
summarized.
Grouping of the papers in Session 2.
Topics
1. Construction methods
with case studies
1.1 Bored tunnel by
TBM (shield
tunneling)
1.2 Shotcrete method
(mountain tunneling
method, NATM)
2. Ground Treatment
2.1 Ground freezing
2.2 Grouting
3. Load and pressure

3.1 Lining pressure
CONSTRUCTION METHOD WITH

CASE STUDIES
3.2 Pressure on TBM
2.1 Bored tunnel by TBM

More and more practices of bored tunnels by TBM
bring forward more and more requirements for shield
tunnel. Table 2 displays the current trend of development of shield tunnel based on the requirements from
the world market of tunneling.
To meet these requirements, technologies of TBM
are also developed at the same time. The recent
4. Conditioning
additives for EPB
5. Ohters
99
Num. of
papers
Authors
6 papers
(3)
Bakker & Bezuijen

(A, B) He et al.
(3)
Sfriso Guatteri et al.

Fillibeck & Vogt
5 papers
(2)
Hu & Pi
Fillibeck & Vogt
(4)
Guatteri et al.
Bezuijen & vanTol
Gafar et al
Fillibeck & Vogt
7 papers
(5)
Hashimoto et al.
Talmon & Bezuijen
Talmon et al.
Bakker & Bezuijen
(A, B)
(4)
Bezuijen & Bakker
Song & Zhou
Bakker & Bezuijen
(A, B)
2 papers Hajialilue-Bonab
et al. (A, B)
3 papers Deng & Zhang
Kuzme & Hrustinec
Li et al.
Table 2.
Current trend in shield tunneling.
Long distance
High speed excavation
Deep excavation
Large cross section
Deformed cross section
High durability of tunnel
Cost performance
3 km10 km
300 m1000 m/month
40 m100 m
10 m15 m of diameter
2 faces4 faces, non-circular
100 years
Not cheap but high quality
with reasonable cost
Table 3. Geotechnical aspects for bored tunneling (shield

tunneling).
TBM type
Applicable ground
Ground loss
Face stability
Filling tail void
Segmental lining
Figure 1. Support pressures before, during and after the

Blow out at the 2nd Heinenoord tunnel (by Bakker &
Bezuijen (A)).
Both of slurry and EPB type in the

soft soil with ground water
Soft to stiff clay, loose to dense sand,
gravel
Possible to be controlled less than
0.11% in normal condition
Need some controlling technologies
for each slurry type or EPB type
Simultaneous grouting can reduce
ground loss and give an uniform
distribution of lining pressure
Many types of segmental lining have
been developed
development of TBM and its technologies are shown

as following:
Figure 2. Surface settlements; measured and backcalculated with different material models (by Bakker &
Bezuijen (A)).
Durability of TBM
Durability of cutter bits
Exchangeable cutter bits
Installation of linings, new segmental linings
Driving control system
Docking method
Backfill grouting
In the practice of bored tunneling by TBM, the

geotechnical aspects shown in Table 3 are of the most
importance and should be well considered.
Bakker & Bezuijen (A, B) shared their invaluable
experiences and findings on shield tunneling in soft
ground obtained in last ten years. During the construction of the 2nd Heinenoord Tunnel that is approximately in the middle underneath the river Oude Maas
in the Netherlands. They found out that because blowout occurred during TBM driving under the river,
face support pressure dropped within 15 seconds after
the cutter face working, shown as Figure 1. According
to their investigation, they pointed out that face support
pressure should be controlled between lower and upper
limits for situations with little overburden or the soil
cover itself is relatively light. We also are interested in
the 15 seconds, which indicted that the front instability occured without any omen, a careful control of
front pressure is necessary. Some analysis results of
surface settlement were also displayed, shown in Figure 2. It was concluded that for an adequate prediction
of deformations it is important to model the grouting
pressure as a boundary condition, in combination with
the use of small strain material model.
As to the structure issues of the 2nd Heinenoord
Tunnel, Bakker & Bezuijen (B) investigated the cracking and palling that occurred due to construction load,
see Figure 3. Then a large scale tunnel ring tests was
carried out, shown as Figure 4. By combining the
model tests as well as numerical tests, it was found
that the usage of kaubit in the ring joint was the main
reason. The compression of the flexible kaubit strips by
jacking force resulted in a slipping of different segment
piece, leading to local stress concentration and irregular deformation. By replacing it with stiffer plywood
plates, the damage was prevented. The influence of the
duration of plywood to the long-term behavior of tunnel, however, is still questionable. During construction
of the first tube for the Westernscheldt Tunnel, they
found out that high grout pressures and in absence of
bedding may cause the buckling of theTBM. Certainly,
100
Figure 3. Damage to the dowel and notch sockets during

the first 150 m of construction of the 2nd Heinenoord tunnel
(by Bakker & Bezuijen (B)).
Figure 5. The vertical stress increment in 1.5 m ahead of the

opening face (by He et al.).
Figure 4. Large-scale tunnel ring testing (by Bakker &

Bezuijen).
some other factors that were not discussed in the paper

may also cause TBM deformation.
He et al. studied the first application of DOT tunneling in Shanghai. They conducted an in-situ test
to investigate the distribution of stress and displacement around the tunnel. Figure 5 shows the vertical
soil stress increment ahead of cutter face. Beautiful distribution of vertical earth pressure increment
and settlement troughs were observed. It is expected
that more detailed information about the measuring
methods can be given out. They also reported a DOT
shield passed under a five-floor building with a distance of 1 m successfully by careful operation, shown
as Figure 6. The main countermeasures were relative
low advancing speed and extra backfill grouting.
Figure 6. Dot shield tunnel run across the buildings

(by He et al.).
2.2 Shotcrete method (Mountain tunnel

method, NATM)
The geotechnical aspects of shotcrete method (Mountain tunnel method, NATM) are summarized in Table 4.
The design and construction procedures of Metro
tunnels in Buenos Aires from 19982007 were
reported by Sfriso. The characterization of Buenos
Aires soils for tunneling is overconsolidated cemented
101
Table 4. Geotechnical aspects of shotcrete method (mountain tunnel method, NATM).

Face stability
Prelining, face bolt, horizontal

grouting, pipe roof, shotcrete,
et al.
Settlement mitigation
Face stabilization, shotcrete,
foot pile, rockbolt, minibench,
ground improvement ( jet
grouting, chemical grouting,
compensation grouting, et al.)
Geological survey
Geophysical survey (elastic
wave, sonic wave, electric
resistivity, et al.), pilot boring
Monitoring technology
Extensometer, 3D laser
scanning, optical fiber sensing,
digital photogrammetric
system, et al.
Prediction of ground water Lowering of ground water
inflow and preservation
table, subsidence, drying well
of ground water
soil with Nspt > 20, which is very favorable for excavation. As shown in Figure 7, shotcrete tunneling
methods evolved from German method to Belgian
method, and reached an optimal full face excavation. Cut & cover method and underground excavated
method were used for stations. According to the filed
measurement, the surface settlement is in the range
28 mm in general, 415 mm at stations.
Guatteri et al. described the state-of-the-art of
application of ground improvement with all round
(360 ) horizontal jet grouting in Sao Paulo and
Barcelona, shown as Figure 8. Horizontal jet grouting columns were executed around the excavated
section, including the invert, and at the far end of
the conical treatment, to create a watertight chamber. This ground improvement achieves good results
of pre-consolidation, settlement mitigation, reduction
of water flow, and keep of face stability. According to field measurement, ground movements were
controlled within 2030 mm.
Shotcrete excavations with ground freezing, jet
grouting, pipe screen and compressed air supporting
methods were applied in the construction of Munich
Subway. Fillibech & Vogt made a comparison of
different methods of face support in settlement sensitive urban areas based on the ground deformations.
In the case of heading with ground freezing under
important structure, measurements for reducing frost
heaves were taken, namely reducing operation time
and careful temperature control. The recorded vertical displacements in Figure 9 show that a maximum
heave of 35 mm was achieved. The report of jet grouting displayed a large heave due to installation of jet
grouting cover, as shown in Figure 10. Although the
face stability increases, the settlement is not reduced so
Figure 7. Various shotcrete tunneling methods in the construction of Metro tunnels in Buenos Aires (by Sfriso).
much as expected. In the paper, it is pointed out that the

installation of crown supporting measures must lead
to higher safety potential, but it is difficult to judge
whether these special measures are necessary or not.
GROUND TREATMENT
The geotechnical aspects of the two sub-subjects of

ground treatment, ground freezing and grouting, are
summarized in Table 5 and Table 6.
102
ColumnCCPh
Septum
Section Tunnel
ColumnsCCPh
Figure 10. Large heave due to horizontal jet grouting (by

Fillibech & Vogt).
Table 5.
Geotechnical aspects on ground freezing.
Mechanism of freezing
process and evaluation of
laboratory freezing test
Property of frozen soils
Figure 8. Ground improvement with all round horizontal

jet grouting applied in Sao Paulo and Barcelona (by Guatteri
et al.).
Frost heave and thaw

settlement
Application of freezing
method on underground
construction
Table 6.
Ground injection
Material
The work by Bezuijen & vanTol aimed to make clear

against the question why fractures can occur more easily in the field than in model tests with the same W/C
ratio. Starting from a conceptual model that shown in
Figure 11, they demonstrated in an analytical way that
heterogeneity of soil in the field and the stress reduction by the installation of pipes (so-called TAM) and
other causes before injection are main reasons.
Garfa et al. performed a sereies of laboratory scale
grout injection tests in which various factors affecting fracturing of sand were studied. Figure 12 shows
the schematic diagram of the experimental setup. The
experimental results confirmed that fracture initiation
Strength, stiffness, freezing

point, temperature, thermal
conductivity, salinity
consistency, et al.
Laboratory testing, prediction,
countermeasure
Cross passage, docking of
TBMs, launch and arrival of
TBM
Geotechnical aspects on grouting.
Jet grouting
Grouting
method
Figure 9. Vertical ground surface displacement along a

tunnel protected by ground freezing (by Fillibech & Vogt).
Segregation potential, ice lens,

structure of soil, et al.
Evaluation of
improvement
Settlement
control
Uniformity of improved soil, ground

deformation during jet grouting,
applicable ground condition
(boulder, obstacle, et al.)
Chemical grout, micro-cement,
CB, LW, polyurethane, et al.
Penetration grouting,
compaction grouting, double
packer, et al.
Fracturing, compaction effects,
uniformity, strengthening, reduction
of permeability
Compensation grouting
in sand requires some local inhomogeneity around the

injection point, rapid development of a filter cake with
a limited thickness and a grout with low viscosity and
a limited yield stress. Grouts with high w/c (watercement) ratio will exhibit fractures with the formation
of filter cake. If the w/c is low, no fractures will be
formed.
103
Table 7. Geotechnical aspects on lining pressure and pressure acting on TBM.

Lining pressure
for design
Longitudinal
deformation and
bending of tube
Pressure on TBM
Figure 11. Sketch with possible deformation modes of the

injection hole (by Bezuijen & van Tol ).
Lining pressure during construction

and at long term, magnitude and
distribution of pressure, effects of
backfill grouting (grouting pressure
and materials), soil types and ground
condition
Backfill grouting, live load and dead
load distribution, subsoil reaction, et al
Backfill grouting pressure, jack force,
slurry pressure, grease pressure at tail
seal, earth pressure at a face, driving
control of TBM, shape and rigidity of
shield, subsoil reaction, tround
deformation by pressure and load
from TBM
Figure 12. Schematic diagram of the experimental setup (by

Safar et al.).
LOAD AND PRESSURE
Geotechnical aspects on lining pressure and pressure

acting on TBM are summarized in Table 7.
Hashimoto et al. analyzed a series of observed earth
pressure measured by PAD type earth pressure cell in
soft clay, stiff clay and sand respectively. It was found
that the distributions of earth pressure are very uniform
in each kind of soils. In the very soft clay, a large portion of the overburden will act upon the lining, while
in the stiff clay and the dense sand, the magnitude and
distribution of earth pressure also depend largely on
the backfill grouting, shown as Figures 13 & 14.
Talmo & Bezuijen presented a very interesting paper
on the prediction of the ground pressure (lining pressure) based on the flow theory of backfill grout in
combination with the time dependent consolidation of
grout material. The measured results and the predicted
results are shown in Figure 15. In their paper, they displayed the results that the lining pressure drops largely
with time, shown as Figure 16. However, according to
the research by Hashimoto et al. the lining pressure
Figure 13. pv /pv0 vs. 2C/pv0 in clayey ground (by

Hashimoto et al.).
drop is very small.The main differences are considered

to be the grout material and injection methods.
Talmon et al. studied the longitudinal tube bending due to grout pressure. They carried out beam
action calculation using input parameters of the bending moment by TBM jacks, transverse force by TBM,
vertical grout pressure gradient behind TBM, loading diagram and unsupported length of tunnel lining
in TBM, et al. The calculated result fits the observed
ones to some extent, shown as Figure 17.
Bezuijen & Bakker described the interaction
between the slurry from the face and the grout from
the tail, see Figure 18. The pressure distribution along
the longitudinal direction of TBM is calculated theoretically based on the pressure loss (P) due to the
flow. The calculated result is shown in Figure 19. They
104
Figure 16. Drops of lining pressure with time (by Talmo &
Bezuijen).
Figure 14. pv /pv0 vs. SPT-N in sandy ground (by Hashimoto
et al.).
Figure 17. Measured and calculated bending moments

compared (Groene Hart Tunnel, the Netherlands) (by
Talmon et al.).
opining ratio of cutter face. They found that EPSR in

clay is larger than that in cobble and sand.
Figure 15. Meausred and calculated grout pressures (by
Talmo & Bezuijen).
found that P depends on the shear stress of the grout

along the TBM (r ), gap between TBM and ground,
and length increment along the TBM.
Song & Zhou did a research work on the earth
pressure distribution of excavation chamber in EPB
tunneling. According to their work, the total supporting pressure can be composed by two parts: (1) Earth
supporting pressure PE in working chamber; (2) Cutter head plane supporting pressure PP . The authors
proposed an estimation method of earth pressure ratio
CONDITIONING ADDITIVES FOR EPB
Required properties of conditioned soil for EPB are

summarized in Table 8. Until now, there are many
types of conditioning additives have been utilized in
practice, including slurry, foam, polymer, water (for
clayey ground), cellulose, sodium alginate, et al.
There are many data in the world especially in Japan
for this subject, but these data has not been summarized in general. In Session 2, there are two papers
by Hajialiue-Bonal et al. concerning condition additives for EPB. The two papers described the following
results from laboratory test for foam and conditioned
sandy by foam:
based on the empirical relation among cutter head

torque, trapezoidal shape of pressure distribution,
105
1 Polymer type foam shows a good stability;

2 With some combination, foam/sand mixtures have
high compressibility;
Table 8.
Required properties of conditioned soil for EPB.
High flowability
Impermeability
High
compressibility
Uniformity
Low shear strength, reduction of

cutting torque and wear of cutter bits,
stable and precise monitoring of earth
pressure in working chamber
Prevention of water inflow and piping
Reducing earth pressure fluctuation
during excavation
Prevention of water inflow and piping,
precise monitoring of earth pressure
and many new technologies developed in recent

years were introduced. This is particularly important,
because through case study, we can understand the
advantage, disadvantage, applicable field, and feasibility of the new construction methods and technologies.
Some papers introduced new monitoring and measuring technologies. This is also an important aspect
for tunnel construction. On-time and accurate monitoring and measuring can make tunnel construction
work quicker, safer, and more economical. Of course,
abundance of measuring data help us analyze and
understand the mechanism and the essence of the
interaction between tunnel structure and ground.
Prediction and theoretical analysis was concerned
in some papers. In general, prediction and analysis
results were compared with observed results to verify their validity. But we should pay attention to the
limitation and applicable field of these methods.
Figure 18. Possible flow directions and sketched pressure

distributions along the TBM.
REFERENCES
Figure 19. Pressures and joint width along a TBM.
3 The soil conditioning by foam cause decrease of

shear strength (c,
);
4 When Foam Expansion Ratio is 10 < FER < 18, the
change of strength are negligible.
6
CONCLUSIONS
The papers in this session provide abundant case studies to advance our knowledge of tunnel constructions,
Bakker K.J. & Bezuijen A. 10 years of bored tunnelling in

the Netherlands: Part I geotechnical issures.
Bakker K.J. & Bezuijen A. 10 years of bored tunnelling in
the Netherlands: Part II structureal issures.
Bezuijen A. & Bakker K.J. The influence of flow around a
TBM machine.
Bezuijen A. & van Tol A.F. Mechanisms that determine
between fracture and compaction grouting in sand.
Chong H., Li T. & Yan J. The double-o-tube shield tunnel in
Shanghai soil.
Deng Z.G. & Zhang Q.H. Research of non-motor vehicle -rail
transit-tube interchanging transport system pattern.
Fillibeck J. & Vogt. N. Shotcrete excavations for the Munich
subway Comparison of different methods of face support
in settlement sensitive areas.
Gafar K., Soga K., Bezuijen A., Sanders M.P.M. &
van Tol A.F. Fracturing of sand in compensation grouting.
Guatteri G., Koshima A., Lopes R., Ravaglia A. &
Pieroni M.R. Historical cases and use of horizontal
jet grouting solutions with 360 distribution and frontal
septum to consolidate very weak and saturated soils.
Hajialilue-Bonab M., Ahmadi-adli M., Sabetamal H. &
Katebi H. The effects of sample dimension and gradation on shear strength parameters of conditioned soils in
EPBM.
106
Hajialilue-Bonab M., Sabetamal H., Katebi H. & Ahmadiadli. M. Experimental study on compressibility behavior
of foamed sandy soil.
Hashimoto T., Ye G.L., Nagaya J., Konda T. & Ma X.F.
Study on earth pressure acting upon shield tunnel lining
in clayey and sandy ground based on field monitoring.
Hu X.D. & Pi A. Frozen soil properties for cross passage
construction in Shanghai Yangtze River Tunnel.
Kuzma K. The influence of engineering-geological conditions on the construction of radioactive waste dump.
Li Z.X., Han X. & Wang K.S. Critical ventilation velocity in

large cross-section road tunnel fire.
Sfriso A. O. Metro tunnels in Buenos Aires: Design and
onstruction procedures 19982007.
Song T.T. & Zhou S.H. Study on the Earth Pressure Distribution of Excavation Chamber in EPB tunneling.
Talmon A.M. & Bezuijen A. Backfill grouting research at
Groene Hart Tunnel.
Talmon A.M., Bezuijen A. & Hoefsloot F.J. Longitudinal tube
bending due to grout pressures.
107
Physical and numerical modelling

P.L.R. Pang
Geotechnical Engineering Office, Civil Engineering and Development Department,
Government of the Hong Kong Special Administrative Region, HKSAR
ABSTRACT: This General Report has been prepared based on a review of twenty papers submitted to the
session on Physical and numerical modelling related to geotechnical aspects of underground construction.
The papers cover a wide range of model feature types in different materials. The problems studied include
ground/tunnel face stability, ground/tunnel deformation and earth pressures, ground/tunnel-structure interaction,
seismic behaviour, and vehicle fires in a road tunnel. This report highlights and discusses the approaches used
in modelling and presents the key findings. Some remarks are given at the end on the objectives of modelling
and the work of TC28.
INTRODUCTION
20 papers have been submitted to this session (Table 1).

Three of the papers are joint contributions from authors
of two countries.
The papers cover a wide range of model feature
types (Table 2). These include tunnels in clay, sand,
aluminum rods (modelling a granular mass), layered
soils, as well as tunnels in soft or weak rock. There is a
paper on modelling of deep excavations with steppedtwin retaining walls, and a paper on vehicle fires in a
road tunnel.
2
PHYSICAL MODELLING
Eleven papers present results of physical modelling

(Table 3). These include six models at 1 g, a photoelastic model and four centrifuge models.
Table 1.
Geographic distribution of the papers.
Country
Papers
China
China/France
Denmark
France
Italy
Italy/UK
Japan
Japan/UK
Korea
The Netherlands
UK
5
1
1
1
1
1
2
1
4
1
2
20
For the three models that use aluminum rods, the

tests were carried out at 1 g. Numerical modelling was
also carried out to compare with the results of the
1 g tests.
The sand model at 1 g was prepared by compaction
of the sand using a plate vibrator. The compaction
could have created locked-in compaction stresses
on the model braced wall and the adjacent tunnel
thus influencing the model test results. This was not
discussed in the paper.
Two papers present results of modelling of rock tunnel problems using 1 g tests. One used barite powder,
sand and plaster mixed with water, and the other used
concrete bricks to model the soft rock. There was some
discussion on the modelling laws in the papers. While
the conclusions on the qualitative behaviour seem reasonable, and are not unexpected, it is not sure if the
quantitative results are valid at prototype scale where
discontinuities in the rock and the higher stress levels
could influence the magnitude of the deformations.
Table 2.
Feature types covered in the papers.
Feature type
Tunnels in:
(a) Clay
(b) Sand
(c) Aluminum rods/crushed glass
(d) Layered soils
(e) Soft/weak rock
Deep excavation (aluminum rods)
Vehicle fires in road tunnel
109
Papers
5
3
4
3
3
1
1
20
Table 3.
Papers on physical modelling.
Technique and scale
2D/3D
Materials used
Papers
Laboratory 1 g model
(scales: 1/10,
1/19, 1/80)
2D
3
1
1
Photoelastic models
Centrifuge models
(75 g, 100 g, 160 g)
2D
3D
Aluminum rods
Sand
Barite powder/
sand/ plaster
Concrete bricks
Crushed glass
Clay
Sand
Sand overlying
clay
1
1
1
2
1
There are four papers on centrifuge tests. One of

the papers is to study the effects of pile loading on an
existing tunnel in an overconsolidated clay, two are on
centrifuge tests where dry sand was used to construct
the models for studying the interaction mechanisms,
and one on tunnelling in an overconsolidated clay
overlain by sand under the water table.
3
NUMERICAL MODELLING
18 papers present numerical modelling results

(Table 4). Different numerical modelling techniques
were used.
14 out of these 18 papers used either 2D or 3D
codes based on the finite difference method (FDM)
or the finite element method (FEM). Some of the
codes, e.g. CRISP, FLAC and PLAXIS, are well established codes and the 2D versions are commonly used in
current engineering practice. In the analyses, the soil
was modelled either as a linear elastic or an elasticperfectly plastic material with the Mohr-Coulomb or
Drucker Prager failure criterion. Where a comparison
was made, the elastoplastic model performed better
than the elastic model.
In one paper, a slope stability analysis program
SLOPE/W based on the limit equilibrium method
was used to compute the factors of safety of a clay
slope. The results were compared with the results of
FLAC and PLAXIS which used the strength reduction
method. However, no information is given on the theoretical method used (a few options are available in
SLOPE/W such as Janbu, Bishop and Morgenstern &
Price) and the choice of slip surfaces, which could
affect the computed safety factors. Also, no information is given on what the slope deformation and the soil
shear strain were, when the soil strength is reduced for
the factor of safety to approach unity.
A visco-elastic model adopting a nonlinear relationship between the normalized shear modulus (and
damping ratio) and the shear strain amplitude was used
for a 1D ground dynamic shear response analysis.

The code EERA was used for the analysis, the
objective of this study was to calibrate a linear
visco-elastic, effective stress based, constitutive model
for use in coupled 2D dynamic analyses using the
finite element program PLAXIS. The viscous damping was accounted for using the Rayleigh formulation
(Woodward & Griffiths, 1996).
The subloading tij finite element model (developed
by Nakai & Hinokio (2004)) was used in two cases
to provide results for comparing with physical modelling at 1 g which used aluminum rods in the model
tests. The tij model takes into account the influence
of the intermediate principal stress by introducing a
modified stress tij. Also, the subloading concept (proposed by Hashiguchi (1980)) is adopted to model the
influence of soil density. Five of the seven parameters in the tij model are the same as those in the
Cam-clay model, with one more parameter added to
describe the influence of soil density and confining
pressure, and another parameter added to characterize
the shape of the yield surface. Laboratory biaxial tests
were carried out to compare the stress-strain curves
obtained from the finite element program FEMtij2D. In the biaxial tests, shearing of the aluminum
rods, which had low friction angles, induced dilatant
behaviour. The match between the biaxial tests and the
finite element analysis results appears reasonable but
this is up to a shear strain level of about 12% only
(Figure 1).
The Distinct Element code UDEC was used in one
case to compare with the results of large-scale model
tests carried out using concrete bricks to model rock.
However, the paper does not indicate how the discontinuities in the rock were modelled. For the other two
papers on tunnels in rock, the numerical simulations
were carried out using finite element codes adopting
an elastoplastic rock model with the Drucker-Prager
failure criterion. It seems that the need for modelling
the discontinuities that may be present in the rock
was not considered. It is not too clear from the two
papers how the rock parameters were determined for
the continuum models and the field prototypes.
Results obtained from closed form solutions
derived using upper bound limit analysis were presented in two of the papers, for comparison with
the results of centrifuge modelling and numerical
modelling respectively.
The Fire Dynamics Simulator code incorporating
a large eddy simulation model was used to carry out
computational fluid dynamics modelling. The objective of this work was to study the heat release rates
from vehicle fires in a road tunnel of 15 m in diameter.
The computed results were compared with an empirical equation. This indicates that the empirical equation
requires improvement for the case of small fires in road
tunnels with a large cross section.
110
Table 4.
Papers on numerical modelling.
Constitutive law
Modelling
Program
Papers
Linear elastic
Nonlinear visco-elastic
Elasto-plastic
(Mohr Coulomb)
2D FEM
1D shear
2D FEM
2D FDM
3D FEM
3D FDM
3D FEM
PLAXIS
EERA
CRISP, Msc.MARC
FLAC
MIDAS-GTS
FLAC3D
MARC
2
1
2
1
1
3
2
2D FEM
FEMij-2D
2D DEM
Limit analysis
CFD
UDEC
Closed form solution
Fire Dynamics Simulator
1
2
1
Elasto-plastic
(Drucker Prager)
Elasto-plastic (Cam
clay + 2 parameters)
Distinct element
Rigid-plastic
Large eddy simulation
Figure 1. Stress-strain-dilatancy relation.
PROBLEMS STUDIED
The problems studied as reported in the papers include:

1. ground/tunnel face stability (5 papers),
2. ground/tunnel deformation and earth pressures (8
papers),
3. ground/tunnel-structure interaction (5 papers),
4. seismic behaviour (1 paper), and
5. vehicle fires in a road tunnel (1 paper).
A brief review of selected papers is given below.
4.1
Ground/tunnel face stability
The subject of face stability is a very important one.

If the face pressure applied is too low, there could be
a collapse or excessive ground settlement, and if the
face pressure is too high, there could be a blow-out
failure or excessive ground heave.
A number of researchers have studied this problem (e.g. Anagnostou & Kovri, 1994). The following
papers have added to the knowledge base.
Li et al investigated the failure of a large slurry

shield-driven tunnel using upper bound limit analysis and numerical modelling. The study is for the
15.43 m diameter Shanghai Yangtze River Tunnel constructed in soft clay. A shallow ground cover section,
with a ground cover to tunnel diameter (C/D) ratio
of 0.7, was selected for the study. Undrained conditions were assumed in the modelling. A multi-block
failure mechanism with a uniform face pressure (suggested by Soubra, 2002) was used for the limit analysis.
FLAC3D was used for the numerical modelling (which
adopted an elastic-perfectly plastic constitutive model
with a Mohr-Coulomb failure criterion). The results of
the upper bound limit analysis and the 3D numerical
modelling showed that partial blow-out failure of the
upper part of the tunnel face occurs when the slurry
pressure is large, whereas global collapse of the whole
tunnel face occurs when the slurry pressure is small
(Figure 2).
The authors noted that the difference between the
slurry pressure and earth pressure at the crown and
invert for a large diameter slurry TBM tunnel can be
large and this could have a significant effect on the
failure mechanism and the critical slurry pressure. The
failure mechanisms and the critical slurry pressures at
the tunnel axis level obtained from the limit analysis and the numerical modelling agree well with each
other (Figure 3).
Caporaletti et al reviewed the past research on tunnel stability in undrained conditions (Davis, et al,
1980; Kimura & Mair, 1981; Sloan & Assadi, 1992),
in drained conditions (Atkinson & Potts, 1977) and
in layered ground (Grant & Taylor, 2000). They conducted centrifuge tests to investigate the stability of a
circular tunnel in layered ground, with clay overlain
by a medium dense sandy layer, below the water table.
The C/D ratio of the tunnel was 2.38. The clay was consolidated from a slurry, to give an overconsolidation
ratio ranging between 1.4 and 2.8 with depth. All tests
111
Figure 2. Two kinds of partial failure mechanisms.
were carried out at 160 g. The condition of tunnel collapse was taken as volume loss greater than 20%. In
the centrifuge tests the mechanism of failure for the
layered ground involved a wide area of soil both in
sand and in clay, with pseudo-vertical settlements at
the sand-clay interface (Figure 4).
It was found that the contribution to stability due to
friction acting within the upper sand layer represented
a significant contribution. A significant overestimate
of the tunnel support pressures to prevent collapse
might result if the theoretical solutions obtained for
homogenous clays are used with the sand layer treated
as a surcharge. The authors proposed a new failure
mechanism which provided an upper bound to the
experimental data obtained (Figure 5). It would be
interesting to examine whether the proposed mechanism is applicable for the case of a loose sand layer.
Date et al carried out a series of centrifuge tests
at 75 g to investigate the ground deformation patterns during excavation of tunnels in dry sand. The
C/D ratio of the model tunnels was one, and some of
the models incorporated reinforcements. The ground
deformation was found to be small even when the
face pressure was reduced to half the initial pressure
of 100 kPa, but once movement started upon further
reduction of the face pressure it increased sharply leading to instantaneous collapse (i.e. a brittle failure).
Figure 3. Comparison of failure mechanisms of Case 2

(velocity contour for FLAC3D analysis).
Figure 4. Mechanism of failure from centrifuge tests

(VL
= 20%).
The model tests without reinforcement collapsed at

a support pressure which agrees with the centrifuge
test results of Chambon & Cort (1994). The study
found that introduction of face bolts and forepoling
yielded different tunnel collapse mechanisms, which
depended on the density of the face bolts and forepoling bolts. Surprisingly, the reinforcements contributed
112
Figure 5. Mechanism of failure for layered ground.
Figure 7. Measured excess pore pressure in front of a slurry

shield and approximation.
Figure 6. Tunnel failure patterns on the longitudinal section.
to only a slight reduction in the support pressure

required to keep the tunnel face stable, compared to the
case without reinforcement. The face bolts installed
stiffened the ground ahead of the face and were found
to be able to reduce the face extrusion. The forepoling divided the ground around the tunnel face into two
zones, with the outer zone forming an arch comprising
the forepoling bolts. The geometries of the collapse
mechanisms are similar to those observed by other
researchers for tunnels in sands, e.g. as reported by
Chambon & Cort (1994) and Mair & Taylor (1997).
They all involve a narrow chimney, propagating
almost vertically from the tunnel up to the ground
surface (Figure 6).
FLAC3D analyses were also carried out. The MohrCoulomb soil model with strain softening/hardening
was found to give a better match to the centrifuge data
than the Mohr-Coulomb model without strain softening/hardening. The deformation pattern obtained from
the analysis for a model reinforced with face bolts was
similar to that of the centrifuge test but the magnitude
was smaller. The authors recommended to study further the effect of mesh shape and the effect of changes
in soil-bolt interaction properties upon excavation in

the numerical analyses.
The information on failure mechanisms presented
in the above papers is interesting and useful. There is
recent improvement in understanding of the groundtunnelling interaction processes associated with influence of grouting pressures, removal of the filter
cake and the pore pressures generated during the
advance of a slurry TBM (Figure 7). This was achieved
through field measurements obtained during construction (Bezuijen & Talmon, 2008). Further data and
study in this area will no doubt augment the results
of existing laboratory and analytical modelling, which
have not accounted for such processes. Further understanding of the processes could help to evaluate the
need to refine the calculation models and design methods for estimation of face pressures required to prevent
collapse and blow-out.
4.2 Ground/tunnel deformation and earth
pressures
A number of papers in this session present results of
modelling to study the ground deformation and earth
pressures around a tunnel.
Shahin et al developed a new circular tunnel apparatus and conducted 1 g model tests to examine the
ground movements induced by tunnelling and the earth
pressures around the tunnels. Aluminum rods were
used to model a granular soil mass. The surface settlement was measured using a laser type displacement
transducer with an accuracy of 0.01 mm, and photographs were taken during the experiments which
were later used as input for the assessment of the
ground movements using the Particle Image Velocimetry technique (White et al, 2003). To compare with
the model test results, numerical simulations were
carried out using 2D finite element analyses under
plane strain and drained conditions. The computer
program FEMtij-2D was used. The initial stresses
113
Figure 8. Distribution of shear strain: tunnel invert is fixed.
applied correspond to the self-weight condition. Two

C/D ratios, viz. 1 and 2, were examined. The effects of
full face excavation (with the centre of the excavation
kept fixed) and top drift excavation (with the invert of
the tunnel kept fixed) were also studied. The surface
settlement and earth pressures around the tunnel were
found to be significantly influenced by the displacement at the tunnel crown for the same overburden and
same volume loss.The volume loss was less significant
compared to the crown drift in the case of the shallow
tunnel. The full face excavation case produced a wider
shear deformation region than that for the case of top
drift excavation (Figure 8). The use of an elastoplastic
soil model produced better match with the model test
surface settlement profile than an elastic soil model.
The distribution of earth pressures around the tunnel
depended on the excavation pattern. The authors indicated that the numerical simulations were generally in
good agreement with the model test results. However,
it is no clear whether the tij finite element model is
capable of describing the behaviour of tunnels constructed in real soils especially in soils which exhibit
contractile behaviour.
Liang et al studied the effects of soil stratification
on tunnelling-induced ground movements. 3D analyses were carried out using the computer program
FLAC3D. The behaviour of the 2.47 m diameter Thunder Bay sewer tunnel in Canada, constructed using a
TBM with segmental concrete lining, in soft to firm
clays with silt and sand seams, was simulated. The C/D
ratio of the tunnel was 3.8. The soil strata were divided
into four sub-layers for the purpose of the analyses.
Figure 9. Lateral displacement 15 m behind the tunnel face.
The ground surface settlement, lateral displacement

profile at 15 m behind the tunnel face and the subsurface settlement with depth above the tunnel axis from
the analyses were compared with the analysis results
obtained by Lee & Rowe (1991) using the FEM3D
program (also based on an elastoplastic soil model).
They were also compared with the field data reported
by Belshaw & Palmer (1978). Additional comparisons
were carried out with the analytical solution given by
Loganathan & Poulos (1998). The study showed that
the elastoplastic soil model could simulate the deformation profiles better than those based on the elastic
model. The results of the elastoplastic soil model indicated that soil stratification had little effects on the
ground surface settlement but significantly influenced
the lateral displacement and subsurface settlement
profiles (Figure 9). This was different from the elastic soil model which predicted that soil stratification
had significant effects in all cases. This is an interesting case history of benchmarking a 3D computer
program using data from a past project, illustrating
the value of documenting good data and making it
available for research.
Song et al studied the time-dependent behaviour
of soft ground tunnels constructed using steel reinforcements grouted into the ground ahead of a tunnel
114
tests showed that the ground displacements induced

by tunnelling were mainly within a zone of 0.25D
from the tunnel, where D is the tunnel width. Horizontal displacements of more than 40% and vertical
displacements more than 20% of the total displacements occurred during excavation of the pilot tunnel.
The authors suggested that the stability of the 2-arch
tunnel could be dominated by the stability of the pilot
tunnel excavation and that the rock bolt length should
be longer than 0.25D. Displacements obtained from
UDEC analyses were presented. While these showed
the same pattern, details of the analyses were not
given. Based on the limited measurements obtained,
the authors suggested that the rock load acting on the
centre pillar of the 2-arch tunnel may be taken to be
0.15W for preliminary design, where W is the centreto-centre distance between the left and right tunnels,
when the RMR of the rock mass is more than 60. No
numerical analyses were carried out. More research
was recommended to confirm the proposed empirical
relationship. It would be useful to examine the influence of rock discontinuities and the effect of rock block
size relative to the tunnel diameter.
4.3
Figure 10. Time-dependent characteristics of elastic wave
velocities of a sand-cement mixture.
(a technique which the authors called the reinforced

protective umbrella method). Laboratory direct shear
tests and P and S wave velocity tests (using piezoelectric bender elements) were carried out to determine
the strength and stiffness of the sand-cement mixture
at different curing times. The test results showed that
the sand-cement mixture gained significant increases
in stiffness after about 6 hours whereas the apparent cohesion increased to about 2 MPa after 7 days
(Figure 10).
3D finite element analyses were carried out using
a computer program MIDAS-GTS (2005) to simulate
the behaviour of such a tunnel. The tunnel is 18.8 m
wide and 10.4 m high, at 15 m below ground. It was
constructed in weathered rock, using 12 m long steel
pipes as reinforcement. The water table was at ground
surface. The analyses incorporated the time-dependent
material properties of the sand-cement mixture. The
excavation rate was taken as 0.75 m per day. The
study concluded that use of the 23 days strength and
stiffness parameters was adequate for predicting the
time-dependent deformation behaviour, for practical
design purposes, provided that there is sufficient overlap between the reinforcements. No comparison with
any field performance monitoring results was however
presented.
Lee et al studied the behavior of a 2-arch rock tunnel using a large-scale test machine (6 m wide 6.5 m
high). The model tests (1/19 scale) were conducted at
1 g. The rock was modelled using concrete bricks. The
Ground/tunnel-structure interaction
Broere & Dijkstra investigated the influence of tunnel

volume loss on piles using the photoelastic technique.
2D plane strain model tests were conducted to examine
the tunnel-pile interaction. Crushed glass (a photoelastic material) was used to model the soil. The effects
of volume loss were simulated by making the tunnel diameter contract vertically. From the tests, it was
found that significant stress changes occurred close
to the pile tips. The tests with a volume loss of 0.6%
showed a clear influence of the volume loss on the
stresses near the pile tips up to one tunnel diameter
away. The study suggested that the influence zone for
displacement piles with both end bearing and skin friction, might be slightly larger than for bored piles with
end bearing alone. The authors indicated that further
field observations, model testing and numerical modelling are required to determine the influence zone.
Lee & Yoo studied the ground shear strain patterns developed around a tunnel and the existing piles
nearby due to tunnel construction. Small-scale laboratory model tests at 1 g were conducted. Aluminum
rods were used to model the soil mass and the piles
embedded in it. A tunnel diameter reduction system
capable of achieving a tunnel volume loss of up to
20% was specially developed. The strained controlled
tests carried out using this system resulted in ground
shear strains which were captured by close range photogrammetry. 3D numerical analyses were also carried
out using the finite element program CRISP. Comparison between the physical model tests and the finite
element analyses showed good agreement in terms of
shear strain patterns. Based on the maximum shear
115
Figure 11. Schematic illustration of shear strain modes for

pile-soil-tunnelling interaction.
strain contours, two distinct shear strain patterns were

observed, viz. with and without tunnel-pile interaction
(Figure 11). The boundary between these two modes
of behaviour depended on the location of the pile tip
from the tunnel and the magnitude of the tunnel volume loss. The authors suggested that this boundary
might serve as a useful guide in the planning the tunnel alignment in areas where piles are present. It may
be worthwhile to compare the results given in these
papers with the findings of Jacobsz et al (2004; 2005)
from centrifuge tests and Selemetas et al (2005) from
field tests.
Yao et al studied the effects of loading of bored
piles on existing tunnels. Centrifuge model tests were
carried out at 100 g. The model tunnel was formed
in firm to stiff clay consolidated from a slurry. The
tunnel lining deformation, pore pressures in the clay,
pile load applied, pile settlements and tunnel face pressures were monitored while the pile loading was being
applied. Two C/D ratios, viz. 2 and 3, were studied. The
tests examined the behaviour after pile construction.
The influence of pile excavation was not considered. In
the tests the pile base was set at two different positions:
tunnel crown and invert level. The rate of loading was
designed to create undrained conditions. Preliminary

analysis of the results indicated that the pile settlement
had a linear relationship with increase in applied load
when the load exceeds half the designed ultimate load.
The tunnel centre always moved downwards and away
from the pile. Increasing the pile-tunnel clear spacing reduced the deformation of the tunnel lining. The
long pile had more effect on the tunnel lining than the
short pile regardless of the C/D ratio. The tunnel crown
was always subject to significant movement due to pile
loading.
Marshall & Mair investigated the soil-structure
interaction mechanisms resulting from tunnel construction beneath buried pipelines using centrifuge
modelling. The study aimed to validate visually the
interaction mechanisms that account for pipeline
behaviour. Particle Image Velocimetry was used to
measure displacements for characterising the soilstructure interaction. The model tests were carried
out at 75 g, using sand prepared to a relative density of 90%. The C/D ratio of the tunnel was 2.4. The
study showed that estimation of the tunnel volume loss
(defined as change in tunnel volume divided by the
original total tunnel volume) using soil displacement
data was not simple for sands. This was due to the
uncertainty on the extent of the dilation and contractile behaviour of the sand around the tunnel. The soil
volume loss (defined as the volume calculated by integrating the soil settlement profile and dividing by the
original total tunnel volume) was not always the same
as the tunnel volume loss. The magnitude of the former
calculated at the ground surface can be greater or less
than the latter. The centrifuge pipeline test illustrated
that a gap formed below the pipeline at a tunnel volume loss of between 1 and 2%. The gap grew as the
tunnel volume loss increased. The bending moments
induced in the pipe increased from the onset of tunnel
volume loss but did not appear to be sensitive to the
growth of the gap height (Figure 12).
Lee & Kim studied the behaviour of a braced excavation in sand adjacent to a tunnel using large-scale
(1/10 scale) model tests at 1 g. The braced wall was
subjected to preloading to limit the wall deflections
during the ground excavation. The tunnel was at a
distance of half the tunnel diameter from the braced
wall. The sand was prepared to a relative density of
56%. 2D numerical analyses were carried out using the
finite element program PLAXIS. It is not clear what
constitutive model was used for the sand. The study
found that if the wall deflections were significantly
reduced by preloading, the stability of the adjacent
tunnel would greatly increase. The maximum bending
moment and shear force in the tunnel lining decreased
due to the preloading. The ground surface settlement
also decreased as a result of preloading. The wall
deflection profiles from the model tests agreed well
with the numerical analysis results. It is noted that the
116
Figure 12. Derivation of bending moments from deformed

shape of pipeline (from PIV data).
sand was compacted to construct the models. However,

it is not clear what initial soil stresses were used in the
numerical analyses. Also, no information is given on
whether the wall installation and excavation sequence
in the analyses matched those in the model tests.
5
CONCLUDING REMARKS
In the papers submitted to this session, the modelling objectives are generally not explicitly stated but
they probably include one or more of the following
objectives:
To observe/understand collapse mechanisms
To observe/understand deformation patterns and
interactive behaviour
To assess/verify the usefulness or accuracy of theoretical solutions, software or empirical rules against
laboratory (1 g or Ng) model test data
Same, but against field measurement data
To benchmark theoretical solutions or software
To predict field performance
Other than to gain knowledge and to understand
the problem, an important goal of the modelling
research should be to provide useful and reliable tools
or to enhance the existing tools for prediction of
field performance, for use in engineering practice. In
this regard, some of the papers have contributed to
this goal.
TC28 has recently set up two working groups, one

on databases on underground works and another on
preparing guidelines for comparing field or physical
modelling with numerical simulations. The first initiative will be useful for modellers in that good quality
data will be archived systematically for easy reference
and retrieval, physical modellers could use the information to plan their research and check their model test
results against otherswork for benchmarking purpose,
and numerical modellers could use the data to check
the reliability and limitations of the existing theoretical closed form solutions and numerical codes. The
second initiative will be useful for those who are carrying out modelling to understand real behaviour or to
validate numerical codes.
One of the databases on underground works could
include failures observed in model tests and failures
in actual projects. Data in the latter category are more
difficult to obtain unless there is a forensic investigation and the information is subsequently made publicly
available. Information on failures would be invaluable to provide lessons learnt, for calibration of design
methods and for providing insights for risk management. If this is to proceed, then there may be merit to
collect data on the size of the failure influence zones
for different ground and geometrical conditions, and
also the time for any cavity created at depth to migrate
to the ground surface. Such information is potentially
useful for risk management, in particular, for preparing
monitoring plans, planning of risk mitigation measures, and preparation of emergency preparedness and
contingency plans.
Another database will be on monitoring results.
For such databases, the monitoring data should be
accompanied by the necessary data on ground and
groundwater conditions, the way the soil and rock
parameters was measured and interpreted, the method
of wall installation, information on ground treatment
and the sequence of construction. Such data would
allow numerical modellers to check the capabilities
and limitations of the existing computer programs.
From a practising engineers point of view, it is often
not practical to use overly sophisticated software
requiring multiple parameters to characterize the soils
for design. This is because of the cost, time and difficulties in obtaining high quality ground investigation
data, the uncertainties associated with modelling the
ground and the hydrogeological conditions (including the boundary conditions), the effort needed to
model the range of design situations and to carry out
sensitivity analyses, the need for having relatively simple tools for undertaking design reviews in a timely
manner during construction, the difficulties in incorporating effectively the wall installation and ground
treatment effects in the analyses, and the lack of competent personnel in the use of sophisticated codes and
checking of the computed results from such codes.
117
There is a lack of systematic comparison on the results

obtained from sophisticated software with those from
less sophisticated ones. The availability of good quality monitoring data and benchmarking of the existing
numerical codes using good quality monitoring data
could help to addresses some of these issues.
TC2 on physical modelling in geotechnics has similar initiatives on databases (see http://www.tc2.civil.
uwa.edu.au). Cross committee communication will
create synergy.
ACKNOWLEDGEMENTS
The author is grateful to Paul Wu for his assistance in
preparing this General Report.
LIST OF PAPERS REVIEWED

Boldini, D. & Amorosi, A. Tunnel behaviour under seismic loads: analysis by means of uncoupled and coupled
approaches.
Broere, W. & Dijkstra, J. Investigating the influence of tunnel
volume loss on piles using photoelastic techniques.
Caporaletti, P., Burghignoli, A., Scarpelli, G. & Taylor, R.N.
Assessment of tunnel stability in layered ground.
Date, K., Mair, R.J. & Soga, K. Reinforcing effects of
forepoling and facebolts in tunneling.
Du, J.H. & Huang, H.W. Mechanical behavior of closely
spaced tunnels laboratory model tests and FEM analyses.
Idris, J., Verdel, T. & Alhieb, M. Stability analysis of masonry
of an old tunnel by numerical modelling and experimental
design.
Iwata, N. Shahin, H.M., Zhang, F., Nakai, T., Niinomi, M. &
Geraldni, Y.D.S. Excavation with stepped-twin retaining
wall: model tests and numerical simulations.
Kasper, T. & Jackson, P.G. Stability of an underwater trench
in marine clay under ocean wave impact.
Lee, S.D., Jeong, K.H., Yang, J.W. & Choi, J.H. A study on
behavior of 2-arch tunnel by a large model experiment.
Lee, S.D. & Kim, I. Behavior of tunnel due to adjacent ground
excavation under the influence of pre-loading on braced
wall.
Lee, Y.J. & Yoo, C.S. Two distinctive shear strain modes for
pile-soil-tunnelling interaction in a granular mass.
Li, Y., Zhang, Z.X., Emeriault, F. & Kastner, R. Stability
analysis of large slurry shield-driven tunnel in soft clay.
Liang, F.Y., Yao, G.S. & Li, J.P. Effects of soil stratification
on the tunneling-induced ground movements.
Marshall, A.M. & Mair, R.J. Centrifuge modeling to investigate soil-structure interaction mechanisms resulting from
tunnel construction beneath buried pipelines.
Shahin, H.M., Nakai, T., Zhang, F., Kikumoto, M.,
Tabata, Y. & Nakahara, E. Ground movement and earth
pressure due to circular tunneling: model tests and numerical simulations.
Song, K.I., Kim, J. & Cho, G.C. Analysis of pre-reinforced
zone in tunnel considering the time-dependent performance.
Wang, K.S., Han, X. & Li, Z.X. Vault temperature of vehicle
fires in large cross-section road tunnel.
Wang, X.M., Huang, H.W. & Xie, X.Y. Effects of different

bench length on the deformation of surrounding rock by
FEM.
Yao, J., Taylor, R.N. & McNamara, A. The effects of loaded
bored piles on existing tunnels.
You, G.M. 3D FEM analysis on ground displacement induced
by curved pipe-jacking construction.
REFERENCES
Anagnostou, G. & Kovri, K. 1994. The face stability of
slurry-shield driven tunnels. Tunnelling and Underground
Space Technology. 9(2), 165174.
Atkinson, J.H. & Potts, D.M. 1977. Stability of a shallow
circular tunnel in cohesionless soil. Gotechnique, 27(2),
203215.
Belshaw, D.J. & Palmer, J.H.L. 1978. Results of a program of
instrumentation involving a precast segmented concretelined tunnel in clay. Canadian Geotechnical Journal, 15,
573583.
Bezuijen, A. & Talmon, A.M. 2008. Processes around a TBM.
Keynote Lecture. Pre-print Volume of the Proceedings of
the 6th International Symposium on Geotechnical Aspects
of Underground Construction in Soft Ground, Shanghai.
Chambon, P. & Cort, J.F. 1994. Shallow tunnels in cohesionless soil: Stability of tunnel face. Journal of Geotechnical and Geoenvironmental Engineering, ASCE, 120(7),
11481165.
Davis, E.H., Gunn, M.J., Mair, R.J. & Seneviratne, N.
1980. The stability of shallow tunnels and underground
openings in cohesive material. Gotechnique, 30(4),
397416.
Grant, R.J. & Taylor, R.N. 2000. Stability of tunnels in clay
with overlying layers of coarse grained soil. Proceedings
of GeoEng2000. Melbourne, Australia.
Hashiguchi, K. 1980. Constitutive equation of elastoplastic
materials with elasto-plastic transition. Journal of Applied
Mechanics, ASME, 102(2), 266272.
Jacobsz, S.W. Standing, J.R., Mair, Soga, K., Hagiwara, T. &
Sugiyama, T. 2004. Centrifuge modelling of tunnelling
near driven piles. Soils and Foundations, 44(1), 5158.
Jacobsz, S.W., Bowers, K.H. and Moss, N.A. 2005. The
effects of tunnelling on pile structures on the CTRL. Preprint Volume of the Proceedings of the 5th International
Symposium on Geotechnical Aspects of Underground
Construction in Soft Ground, Amsterdam.
Kimura, T. & Mair, R.J: 1981. Centrifugal testing of model
tunnels in soft clay. Proceedings of 10th International
Conference on Soil Mechanics and Foundation Engineering. Stockholm, 1, 319322.
Lee, K.M. & Rowe, R.K. 1991. An analysis of threedimensional ground movements: the Thunder Bay tunnel.
Canadian Geotechnical Journal, 28, 2541.
Lee, K.M., Rowe, R.K. & Lo, K.Y. 1992. Subsidence owing
to tunnelling. I. Estimating the gap parameter. Canadian
Geotechnical Journal, 29, 929940.
Loganathan, N. & Poulos, H.G. 1998. Analytical prediction
for tunnelling-induced ground movements in clays. Journal of Geotechnical and Geoenvironmental Engineeing,
ASCE, 124(9), 846856.
Mair, R.J. & Taylor, R.N. 1997. Theme lecture: Bored tunnelling in the urban environment. Proceedings of 19th
118
International Conference on Soil Mechanics and Foundation Engineering, Hamburg, 23532384.

MIDAS-GTS. 2005. Geotechnical & Tunnel Analysis System,
MIDAS Information Technology Co., Ltd.
Nakai, T. & Hinokio, M. 2004. A simple elastoplastic model
for normally and over consolidated soils with unified
material parameters. Soils and Foundations, 44(2), 5370.
Selematas, D., Standing, J.R., & Mair, R.J. 2005. The
response of full-scale piles to tunnelling. Pre-print Volume
of the Proceedings of the 5th International Symposium
on Geotechnical Aspects of Underground Construction in
Soft Ground, Amsterdam.
Sloan, S.W. & Assadi, A. 1992. Stability of shallow tunnels in
soft ground. Predictive Soil Mechanics, Thomas Telford,
London, 1993, 644662.
Soubra,A. H. 2002. Kinematical approach to the face stability

analysis of shallow circular tunnels. Proceedings of 8th
International Symposium on Plasticity, British Columbia,
Canada, 443445.
Wang, J.N. 1993. Seismic Design of Tunnels: A State-ofthe-art Approach. Monograph 7, Parsons, Brinckerhoff,
Quade & Diuglas Inc., New York.
White, D., Take, A. Bolton, M.D. 2003. Soil deformation
measurement using particle image velocimetry (PIV) and
photogrammetry. Gotechnique, 53(7), 619631.
Woodward, P.K. & Griffiths, D.V. 1996. Influence of viscous
damping in the dynamic analysis of an earth dam using
simple constitutive models. Computers and Geotechnics
19(3), 245263.
119
Case histories
A. Sfriso
Department of Estabilidad, University of Buenos Aires, Argentina
ABSTRACT: Twenty papers accepted for publication under IS-Shanghai 2008 Session 3 on Case Histories are
classified and summarized. Papers deal with open pit excavations, ground improvement, tunnels, monitoring
systems and impact on surroundings, most of them related to projects performed in challenging urban environments. It has been found that different authors follow different approaches when reporting case histories, mainly
with respect to the quantitative description of ground conditions and behavior. While this can be attributed to
different scientific and professional schools, it is judged that a higher degree of consistency and completeness
of the basic information is required for a better usability of the informed data. To contribute to this goal, a short
guideline on reporting case histories is proposed.
INTRODUCTION
IS-Shanghai 2008 became a major opportunity to share

recent experience related to underground construction
in soft ground. In Session 3, devoted to case histories, many projects reflecting advances in geotechnical
engineering related to challenging urban conditions
were discussed.
Eight countries contributed a total of twenty papers
to this session: eleven from China, one from Japan,
two from Korea, one from Singapore, one from
Taiwan, one from Italy, two from France and one from
Germany.
In the following sections, the twenty papers are
classified and summarized. The purpose of this classification is to guide the interested reader to specific
information that might be useful for his/her research.
Papers are grouped as follows: i) seven papers dealing with open pit excavations; ii) four papers dealing
with NATM and drill&blast tunnels; iii) five papers
dealing with TBMs and shield tunnels; and iv) four
papers dealing with monitoring systems and the evaluation of the impact of under-ground projects on
surroundings.
Almost all kind of difficult ground conditions due
to existing infrastructure and space constraints were
described. For instance, papers dealing with TBM
tunnelling describe crossing beneath a shield tunnel
and a railway line, across the foundations of a highway bridge and above existing tunnels of an operating
metro line. Open pit excavation projects are not simpler, showing the challenges that urban construction
poses to geoengineering.
Two facts became evident during the revision process, as follows: i) ground conditions are described in
widely different ways and with highly varying degree
of completeness; and ii) ground and structure behavior
are characterized by some representative numbers
selected with ample liberty. While the reported data is
very valuable, some effort must be done to fully exploit
its usability.
It is remarkable that seven out of twenty papers deal
with either recent or on-going underground projects
in Shanghai, thus reflecting the impresive rate of
infrastructure development of the city. A large amount
of information is provided with respect to Shanghai
soils, including laboratory and field tests and extensive reporting of ground behavior during construction.
It is desirable that this valuable information be further analyzed by researchers to produce a consistent
and complete set of material parameters for Shanghai
soils, as the raw data provided by the papers does not
allow for a complete understanding of soil conditions
and soil behavior.
In each section, the list of papers is listed in a
table and a brief description is presented for each
paper.The description merely states the type of project,
geology conditions where known and the description
of a few selected contributions. These contributions
can be of any type, from an overall description of construction processes to a quantitative measure of ground
behavior or detailed monitoring data. The writer
recommends the reading of all papers, as the valuable information contained there is only superficially
grasped by the short descriptions that follow in this
report.
121
Table 1.
Papers on open pit excavations.
Author
Project
Hsiung &
Chuay
Konda et al
Liu, D. et al
Liu, G. et al
Liu, T. et al
Mei et al
Osborne et al
Three excav. 20 m deep in Kaohsiung,

Taiwan
Eleven braced excavations at Osaka, Japan
18 m deep propped excavation in Shanghai
JMM ground treatment in Singapore
Figure 1. Symbol definition (Konda et al 2008).
OPEN PIT EXCAVATIONS

Table 2.
The list of papers dealing with open pit excavations is

given in Table 1.
2.1
Name
Hsiung and Chuay. Observed behaviour of

deep excavations in sand
The behavior of three excavations in Kaohsiung,

Taiwan, is described. The excavations are approx. 20 m
deep, supported by propped diaphragm walls 1.0 m
thick and 36 m long, and excavated in medium dense
silty sand with clay layers (NSPT : 530). The water
table is reported at 3 m to 6 m below ground surface.
The maximum lateral wall displacement h max and
surface settlement v max are reported. Values are normalized by the effective height of the excavation He for
comparison among the three projects. It is observed
that h max /He falls in the range 0.03% to 0.3% and
that v max is about one half of h max . The effect of the
construction sequence and remedial effects to reduce
surface settlements are discussed.
2.2
Konda et al. Measurement of ground

deformations behind braced excavations
The paper reports surface and wall deformations of

braced support systems used at eleven sites of the
Osaka Subway L8 project in Japan. While details of
the support systems are not informed, the geotechnical conditions are reported to vary widely among
the sites analyzed, from gravels to soft clays. Wall
deformation and surface settlements are described by
area indices as shown in Fig. 1. It is concluded that
the ground settlement area As is about 20%30% of
the wall deflection area A for excavations approx.
21 m deep, but can be much higher if consolidation
settlements occur.
2.3
Liu, D. et al. Research on the effect of buried

channels to the differential settlement of
building
The paper deals with the impact of a deep excavation on adyacent structures in Shanghai, China. The
Fill
Clay
Silty clay
Silty clay
Clay
Silty clay
Sandy silt
Silty sand
Description of the soils (Liu, D. et al 2008).

Bottom
level [m] %
2.93
0.33
3.87
11.87
14.87
21.37
35.87
kN/m3
34.6 18.2
43.0 17.3
49.1 16.8
38.9 17.6
34.9 17.9
32.2 18.0
26.3 18.8
Shear pars
e
-
c[kPa]
0.99 21
1.21 13
1.39 14
1.12 16
1.02 15
0.94
4
0.77
1
[ ]
17.5
17.0
11.0
14.0
18.5
29.0
32.0
excavation is 18 m deep, supported by a diaphragm

wall 0.8 m thick and 26 m long with steel supports,
excavated in Shanghai soft clays. Ground conditions
are described as shown in Table 2. The water table is
assumed to be 1 m below ground level.
The maximum lateral wall displacement h max is
reported to be 60 mm, or 0.33% of the excavation
height. Extensive analysis of the settlement behavior of an adyacent building is reported in the paper,
with emphasis on the non-uniform settlement rate during the excavation stage. While the complexity of the
geological conditions is assessed, no data on the compression and permeability parameters of the soft clay
layers is given and the consolidation process is not discussed, despite the fact that the reported settlement of
the building was up to 125 mm.
2.4 Liu, G. et al. Performance of a deep excavation
in soft clay
The deformation behavior a 21 m deep excavation in
Shanghai, China, is described. The geologic profile is
the quaternary soft alluvial and marine clay deposit
typical of Shanghai City. The authors brief on the
geology and present Fig. 2 that allows for a first understanding of the site conditions. GWL is reported to be
1 m below ground surface.
122
Figure 4. Procedure used to clean the lateral surface of the

last panel (Lui, T. et al 2008).
Figure 2. Soil profile and parameters in Shanghai (Liu, G.

et al 2008).
construction stages. The max lateral wall deflection

was h max = 50 mm, or 0.12% of the total height of the
excavation. Surface settlements are reported but not
associated by the authors to a consolidation process.
Moreover, neither a set of compression parameters nor
an analysis of compression behavior of Shanghai clays
is reported.
Figure 3. The two-drill-one grab method for the construction of the diaphragm wall (Lui, T. et al 2008).
The support system is a 0.8 m 40 m propped

diaphragm wall with base compensation grouting and
prestressed struts. The maximum lateral wall displacement h max is reported to be 55 mm, or 0.32% of the
excavation height. This result is compared with other
measured values in Shanghai and other sites having
rather similar soil conditions.
The effect of the stiffness of multi-propped support
systems is analysed and the three dimensional behavior
of the excavation is assessed. It is concluded that a corner effect exists that reduces the lateral wall displacement corner-to-center ratio h max (corner) /h max (center) to
about 0.390.74.
2.5 Liu, T. et al. The construction and field

monitoring of a deep excavation in soft soils
The paper describes the construction procedure of a
very large and deep excavation performed in Shanghai
clays. The excavation was 263 m long, 23 m wide and
38 m41 m deep, supported by a 1.2 m thick and 65 m
long multi-propped diaphragm wall.
The deep diaphragm wall construction procedure is
described in detail, including the so called two-drillone grab construction method shown in Fig. 3 and the
employment of a counterweight to better clean the last
panels lateral surface before pouring the next panel,
as shown in Fig. 4.
Jet grouting was extensively employed to improve
soil conditions in the passive zone. Reported inclinomenter data shows wall behavior along the
2.6 Mei et al. Excavation entirely on subway

tunnels in the central area of the Peoples
Square
The design and construction of a shallow excavation
6 m deep in Shanghai, China, is described. The particular challenge of this project was that the bottom
of the excavation was placed 3 m above existing tunnels. A support system consisting in a soil-cement
pile wall 3.2 m thick with directional jet grouting was
designed and passive tension piles were provided to
control uplift.
2.7 Osborne et al. The benefits of hybrid ground
treatment in significantly reducing wall
movement: a Singapore case history
The paper reports the first major use in Singapore
of a hybrid ground improvement procedure called Jet
Mechanical Mixing (JMM). JMM is a large diameter deep mixing method that forms a central core of
mixed soil combined with a jet-grouted outer annulus.
A schematic diagram of the drilling tool is shown in
Fig. 5.
The system was employed in the Nicoll Highway
Station Project. The excavation was 27 m deep, supported by a 1.5 m thick and 51 m long diaphragm wall.
The JMM was used to make a base plug of improved
soil 7 m thick below the excavation. Ground conditions
include fill, fluvial sand, fluvial clay and normally consolidated marine clay. The authors concluded that the
ground improvement technique employed reduced the
lateral wall displacements h max by a factor of three.
123
Table 4.
Comparison of wall deformation data.
Author
Soil
He
m
h max /He
%
Hsiung & Chuay

Konda et al
Liu, D. et al
Liu, G. et al
Liu, T. et al
Osborne et al
Sand
Varies
Clay
Clay
Clay
Clay
20
21
18
20
40
27
0.030.30
0.100.24
0.33
0.32
0.12
0.09
2.8
Comparison between wall deformation data
NATM/DRILL&BLAST TUNNELS
The list of papers dealing with NATM and drill&blast

tunnels is given in Table 4.
3.1
Description
EclaircyCaudron et al
Guiloux et al
Quick et al
Yoo et al
Drill&blast tunnel in Bois de Peu, Fr.

Drill&blast tunnel in Morocco
NATM tunnel in Mainz, Germany
Subsidence due to water drawdown
3.2 Guiloux et al. Case history on a railway tunnel

in soft rock (Morocco)
Table 3 lists the lateral wall displacement h max as a

fraction of the excavation height He for the different
projects and construction procedures described. No
correlation can be observed between He and h max ,
confirming the well-known fact that wall deflection
heavily depends on the particular construction procedure, to the extent that it might be concluded
that diaphragm walls and construction procedures are
designed to accomplish lateral wall deflections that
balance the performance requirements of engineers
and clients.
3
Author
clays and soft rocks under 8 m to 140 m of overburen.

The support system was formed by shotcrete, steel ribs
and radial bolts. Unfavourable ground conditions in
the clay soils demanded the use of a structural invert,
forepoling and face bolting.
An interactive design and construction procedure
was employed, where monitoring data was used in an
adaptive design process. The paper reports the construction sequence, the use of monitoring information
to adjust design, and extensive data on face displacements measured at four instrumented sections. It is
shown that extrusion extended one diameter ahead of
the tunnel face and that high face extrusion proved to
be a good indicator of poor ground performance and
risk of face failure.
Figure 5. Schematic diagram of the drilling tool showing

the mixing arm of the JMM machine (Osborne et al 2008).
Table 3.
Papers on drill&blast and NATM tunnels.
Eclaircy-Caudron. Displacements and stresses

induced by a et al tunnel excavation: case of
Bois de Peu (France)
The paper describes the ground response during the

construction of the two twin drill&blast tunnels in Bois
de Peu, France. The tunnels have a cross section area of
130 m2 , a length of 520 m, and were excavated through
The construction process of the Ras RMel tunnel

in Morocco is described. The tunnel is 2.6 km long
and has a cross section of 60 m2 . It was excavated
through weak flysch under 50 m to 150 m overburden
by drill&blast method. The support system consists in
23 cm of shotcrete and steel ribs. A particular feature
of the construction procedure is the use of a formwork
to reduce shotcrete loss, as shown in Fig. 6. Stress-toUCS ratios up to 3.5 and convergences up to 300 mm
were reported, values higher than usual for drill&blast
tunnels in rock.
3.3 Quick et al. Challenging urban tunnelling
projects in soft soil conditions
The design and construction of the New Mainz Tunnel in Germany is presented. New Mainz Tunnel runs
parallel to Old Mainz Tunnel, built in 1884, with a
clearance of 4 m to 50 meters. The tunnel is 1250 m
long, with a cross section of 140 m2 and runs below
buildings with 10 m to 23 m overburden. Soils are
marly clays, silts and sand, and the support system
is a complex combination of bolting, umbrellas, face
bolting and reinforced shotcrete. Ground improvement
techniques employed at some sections to reduce surface settlements are described. It is reported that a
reduction of settlements from 11 cm to 1.5 cm2.5 cm
was achieved by ground improvement.
124
Table 5.
Papers on shield tunnels and TBMs.
Author
Description
Antiga&
Chiorboli
Gong&Zhou
Wang et al
Wong et al
Xu et al
EPB tunnels in Milano, Italy

Tunnel beneath railway line in Shanghai
Crossing below existing tunnel in Shanghai
Crossing above existing tunnel in HKSAR
Crossing bridge foundations, Shanghai
of geotechnical engineering that is accounted for by

consolidation theory and that is simulated with constitutive equations that account for inelastic compression.
The Mohr-Coulomb constitutive model reported to be
used in the model, however, neither simulates inelastic
compression nor includes compression parameters.
Figure 6. Construction of Ras RMel tunnel (Guiloux et al

2008).
SHIELD TUNNELS AND TBMS
The list of papers dealing with shield tunnels and

TBMs is given in Table 5.
Figure 7. Construction procedure for Old and New Mainz

Tunnels (Quick et al 2008).
It is very interesting to note the differences and

similitudes in construction procedures used in two
similar tunnels separated in time by one century, as
shown in Fig. 7.
3.4 Yoo et al. Characteristics of tunneling-induced
ground settlement in groundwater drawdown
environment
The paper studies the effect on surface settlements
of groundwater drawdown due to tunnel construction.
The case analyzed consists in a 70 m2 tunnel excavated through weathered granite with 20 m to 30 m
overburden formed by fill, alluvium and weathered
rock. The support system consisted in pre-grouting,
pipe umbrellas, rock bolts and shotcrete.
Water drawdown produced surface settlements that
started approximately six diameters ahead of tunnel
face and that stabilized six diameters behind it. The
problem was analyzed by a parametric study using a
2D finite element model with Mohr-Coulomb constitutive model. It was concluded that surface settlements
due to tunnel construction have different patterns when
ground-water drawdown is present, when compared to
the normal case.
It must be noted that surface settlements due to
groundwater drawdown are a well-known problem
4.1 Antiga and Chiorboli. Tunnel face stability and

settlement control using earth pressure balance
shield in cohesionless soil
The paper analyzes and compares two case histories of
EPB tunnels driven in Milano, Italy. Both tunnels were
excavated through 50 m to 60 m of medium-dense to
dense alluvial sands and gravels.
The authors provide a comprehensive summary of
factors affecting subsidence of shield tunnels in sands.
They conclude that a high advance rate produces less
volume loss and reduces subsidence and show that
EPB face pressure is poorly correlated to surface
settlements but depends on technological aspects of
backfilling operations.
4.2 Gong and Zhou. Shield tunneling beneath
existing railway line in soft ground
The design and construction of the Metro Line 11
tunnel running below the Hu-Ning railway line in
Shanghai, China is described. The tunnel was driven
through Shanghai soft clays below 11 m overburden.
Water level is reported to be 1 m below ground surface.
The tunnel has a cross section of 30 m2 and is supported by a segmental lining 35 cm thick. Extensive
soil improvement, including jet-piles and grouting,
was performed to control surface settlements. It is
reported that settlements in the improved sectors were
85% lower than those of the unimproved sectors. Fig. 8
shows the longitudinal irregularity of the tracks after
the tunnel was driven below the railway.
125
Figure 8. Longitudinal irregularity of the tracks by

tunneling-induced deformation (Gong and Zhou 2008).
4.3 Wang et al. Supervision and protection of

Shanghai Mass Rapid Line 4 shield tunneling
across the adjacent operating metro line
Figure 9. Horizontal umbrella placed between the new and

existing tunnels (Wong et al 2008).
The paper reports the design, construction and monitoring procedure of the crossing of Mass Rapid Line 4
(MRL4) shield tunnels below existing L2 Metro tunnels in Shanghai, China. MRL4 tunnels have a cross
section of 32 m2 and were driven in Shanghai soft clays
only one meter beneath L2 tunnels at a small crossing
angle. No information is provided with respect to soil
parameters and support systems of both the existing
and new tunnels.
Ground control measures taken to reduce displacements in the existing tunnels are described. It is
remarkable that, despite the short distance between
the new and existing tunnels, control measures did not
include ground improvement due to lack of surface
space.
Extensive monitoring data was generated during the
operation, and some of it is summarized in the paper.
It was found that a strict control of shield deviation,
careful tail grouting and a slow advance rate aided in
controlling L2 tunnel displacements. Shield tail grouting is reported to have influenced settlements some ten
to fifteen meters above and behind the grouting section, depending on grout pressure and grouting stages.
As a general conclusion, authors recommend very slow
advance rates to minimize tunnel induced subsidence.
and ground control measures had to be undertaken.

These measures included extensive jet-grouting of the
whole area and the installation of a physical barrier made by an umbrella of horizontal pipe piles
placed below the new tunnel as shown in Fig. 9.
While a sophisticated monitoring system is reported to
have been installed, there is no information of ground
or tunnel behavior during the construction of MTR
Crossing.
4.5 Xu et al. Application of pile underpinning
technology on shield machine crossing through
pile foundations of road bridge
The paper deals with the design problem of a 39 m2
EPB tunnel hiting 14 piles of a bridge foundation.
The tunnel belongs to Metro Line 10 in Shanghai,
China. Soil conditions are described as fill, organic
soil, clayley silts and clays. Two underpinning schemes
are proposed in the paper: i) the bridges foundations
be reinforced before eliminating the existing piles;
and ii) the existing piles be replaced after foundation reinforcement. It is unclear whether the project
is completed or not.
5
4.4 Wong et al. Kowloon Southern Link TBM

crossing over MTR Tsuen Wan Line tunnels in
HKSAR
The paper describes the construction of Mass Transit
Railway (MTR) Crossing. MTR Crossing is the point
where the new Kowloon S. Link twin tunnels cross
(2 m above) the existing MTR tunnels in Hong Kong.
Kowloon S. Link tunnels have a cross section of 51 m2
and were driven through decomposed and sound granite under 6.8 m overburden by a shield-slurry TBM
with an horizontal clearance of 900 mm. Water table
is reported to be 2.5 m below ground level.
No restrictions to service of MTR tunnels were
allowed, and therefore a series of ground improvement
MONITORING AND IMPACT TO

SURROUNDINGS
The list of papers dealing with monitoring systems and

impact to surroundings is given in Table 6.
5.1 Kim et al. Environmental problems of

groundwater around the longest expressway
tunnel in Korea
A 3D hydrogeologic model that simulates the impact
of Inje tunnel on grounwater level is presented. Inje
tunnel is in fact two 14.5 m wide by 11 km long
drill&blast twin tunnels, claimed by the authors to
form the longest expressway in Korea. The geologic
126
capable of managing high precision monitoring data,

is also described.
Table 6. Papers on monitoring systems and impact to

sourroundings.
Author
Description
Kim et al
Hydrogeological model for Inje Tunnel, Korea
Liu & Wang Description of deformation monitoring systems
Qiu et al
Monitoring system applied at Beijing
Subway L1
Zhao et al
Math. model of settlement induced lining stress
5.3 Qiu et al. 3D deformation monitoring of

subway tunnel
The paper describes the application of LIDAR technology to Beijing Subway L1 tunnel. LIDAR technology allows for a rate of 3D data acquisition of
100.000 points/sec by 3D laser scanning. With this
technology, expensive reflecting prisms do not need to
be used and can be replaced by reflection sheets placed
on the tunnel walls. NURBS (non uniform rational B
splines) technology is used to interpolate the obtained
data, and a mathematical model is developed for the
analysis of the information. An example is given where
a differential settlement of 0.29 mm could be measured
using this technology.
5.4 Zhao et al. Effect of long-term settlement on
longitudinal mechanical performance of tunnel
in soft soil
Figure 10. 3D model of fracture network and tunnel for

steady state simulation of groundwater flow (Kim et al 2008).
profiles is composed by methamorphic rocks with

some superficial debris.
A MODFLOW continuous model was implemented
for the far field simulation of groundwater flow, while
a MAFIC discontinuous model was developed for
the near field model including the effect of grouting
on water flow, as shown in Fig. 10. The conclusion
is that grouting might reduce groundwater inflow
to the tunnel by 53% to 3.6 m3 /h/km, and that the
expected drawdown might be reduced by 65% to about
0.61.1 meters.
5.2
The paper presents a structural model for the induced

longitudinal stress developed in a segmental tunnel
due to nonuniform settlements and apply the theory to
a case history.
As reported by the authors, an unspecified highway
shield tunnel 30 yr old, settled/heaved up to 30 mm in
the last 10 years. The mathematical model was used to
evaluate the structural performance of the tunnel based
on longitudinal curvature radius R. It is concluded that
R < 27300 m may induce leakage; stresses induced by
R < 18800 m may fail segments; bolt yielding should
be expected for R < 15000 m; and that tensile failure
would probably occur for R < 302 m.
Liu and Wang. Deformation monitoring during

construction of subway tunnels in soft ground
The objectives, methods and required precision for

open field and urban tunnel ground monitoring are
discussed in this paper. Tunnel Profile Scanners are
introduced. These are two digital cameras mounted on
a rigid frame that produce a stereoscopic digital image
of the tunnel surface.
A high amount of low-precision displacement data
is recorded, with an estimated error of 5 mm. Statistical
analysis of this data, however, is reported to be useful
as a monitoring tool. An automatic deformation device
using an advanced Geodetic Monitoring Software,
127
SUMMARY OF CASE HISTORIES RELATED

TO PROJECTS IN SHANGHAI
Seven out of the twenty case histories presented at

Session 3 are related to challenging underground
projects in Shanghai, China. This is an unique opportunity to advance in technology and to calibrate design
procedures for soft soils with valuable experimental
evidence.
However, it must be noted that no complete and consistent description of Shanghai soils has been found
among all papers. There is almost no information on
basic index parameters like liquid limit, compression
or recompression indexes and apparent OCR due to
ageing.
The shear strength parameters as reported by the
different authors are listed in Table 7 and can be used
as an example to further illustrate the observed lack
of information. While parameters listed in Table 7 are
Table 7.
Strength parameters reported for Shanghai clays.

Depth: 1015 m
Author
Gong&Zhou
Liu, D. et al
Liu, G. et al
Liu, T. et al
Mei et al
Wang et al
Xu et al
c
kPa
8
14
24.0
11.0
Depth: 2025 m
4.
c
kPa
45
15
15.0
18.5
Fig. 2.
32.0
43
no data
no data
no data
5.
15.5
not and cannot be either undrained or effective

stress parameters, none of the authors explained what
these parameters actually mean.
Shanghai clays are normally consolidated clays
with some degree of ageing. Strength of these soils is
universally characterized by undrained shear strength
su , either determined by in situ testing or lab testing. While it might be argued that there are many
undrained shear strengths for a given clay due to stresspath dependency of shear strength, it must be accepted
that the geotechnical community would appreciate any
reported value of shear strength. The overall lack of a
complete description of Shanghai clays highlights the
value of Fig. 2.
Drained shear strength parameters are probably
fewer in quantity, though also presumed to be widely
available, given the impressive pace of city growth
and the excellent degree of geoengineering involved.
Unfortunately, no clear discussion on the critical state
friction angle of Shanghai soils was found among all
papers.
SHORT, DRAFT GUIDELINE ON

REPORTING CASE HISTORIES
After the experience of reviewing papers submitted

to Session 3 to write this report on Case Histories,
the writer believes that a short and draft guideline on
reporting case histories for projects dealing with soft
soils might be useful. The guideline neither intends
to be complete nor definitive, as it is only based on
the information searched but not found in the papers
during the process of writing this report.
1. Name of the project and location.
2. State of the project by the time of submittal of
the paper: design phase, under construction, or
completed.
3. Basic information on geometry: i) for tunnels,
length, area and a picture showing the cross section
with main dimensions; ii) for excavations, type,
6.
7.
8.
9.
10.
dimensions, structural description of the support

system and a picture showing the support system
and soil profile.
A description of geological/geotechnical ground
conditions and water table.
A comprehensive set of clearly defined soil parameters. Ideally, SPT and/or CPT profiles should
be included. Both total stress and effective stress
shear parameters should be indicated, either as
measured or estimated values. If other strength
parameters are also reported, their meaning should
be fully explained. For problems involving large
subsidence or other compression-driven phenomena, compression parameters and material permeability should also be indicated.
A brief description of the construction process.
Description of ground behavior and unexpected
changes in ground conditions during construction
activities.
Monitoring information when available, or a statement otherwise. Some amount of basic raw data
should be included to better understand and use
some derived parameters like A , see Fig. 1. Fig 8
is a good example of information relevant to the
subject being discussed.
For unusal equipments or construction procedures,
some figures/pictures that better explain the idea,
see Figs. 3 to 7 and 9.
For non conventional calculations and models, an
illustrative picture, see Fig. 10.
In all cases, the source and degree of confidence of

the provided information should be assessed.
A good example of reporting a case history can be
found in a paper by Shao and Macari (Shao and Macari
2008), selected because it deals with a deep excavation
in Shanghai clays. Twenty three key parameters identify each of the six main layers that form Shanghai
soils profile, including water content and void ratio,
classification data, shear and compression parameters,
permeability and SPT values (Shao and Macari 2008).
CONCLUSIONS
Session 3 of IS-Shanghai 2008 became an excellent

opportunity to share experience related to underground construction in soft ground in challenging
urban conditions.
Twenty papers from eight countries, dealing with
open pit excavations, NATM and drill&blast tunnels,
TBMs and shield tunnels, and monitoring systems
were classified and summarized in this report. While
the reported data is very valuable, some effort must be
done to fully exploit its usability because no consistent
procedure was followed by the authors to report ground
conditions and ground behavior during construction
activities.
128
A large amount of information was provided with

respect to Shanghai soils, including laboratory, field
tests and ground behavior during construction. Lack of
definition of the reported parameters is judged to maje
the interpretation of the reported information not easy.
To allow for a better consistency and completeness of
reported data, a short draft guideline on reporting case
histories is proposed.
ACKNOWLEDGEMENTS
The writer wishes to acknowedge the authors of the
summarized papers for sharing valuable information
with the geo-community and the organizing comittee
for inviting him to deliver this general report.
REFERENCES
Antiga, A. and Chiorboli, M. 2008. Tunnel face stability and
settlement control using earth pressure balance shield in
cohesionless soil. IS-039.
Eclaircy-Caudron, S., Dias, D. and Kastner, R. 2008. Displacements and stresses induced by a tunnel excavation:
case of Bois de Peu (France). IS-107.
Gong, Q. and Zhou, S. 2008. Shield tunneling beneath
existing railway line in soft ground. IS-013.
Guiloux, A., Le Bissonnais, H., Marlinge, J., Thiebault, H.,
Ryckaert, J., Viel, G., Lanquette, F., Erridaoui, A. and
Hu, M. 2008. Case history on a railway tunnel in soft
rock (Morocco). IS-367.
Hsiung, B. and Chuay, H. 2008. Observed behaviour of deep
excavations in sand. IS-005.
Kim, S.,Yang, H. andYoon, S. 2008. Environmental problems
of groundwater around the longest expressway tunnel in
Korea. IS-087.
Konda, T., Ota, H., Yanagawa, T. and Hashimoto, A. 2008.
Measurements of ground deformations behind braced
excavations. IS-337.
Liu, D., Wang, R. and Liu, G. 2008. Research on the effect of

buried channels to the differential settlement of building.
IS-118.
Liu, G., Jiang, J. and Ng, C. 2008. Performance of a deep
excavation in soft clay. IS-082.
Liu, S. and Wang, Z. 2008. Deformation monitoring during
construction of subway tunnels in soft ground. IS-120.
Liu, T., Liu, G. and Ng, C. 2008. The construction and field
monitoring of a deep excavation in soft soils. IS-029.
Mei, Y., Jiang, X., Zhu, Y. and Qiao, H. 2008. Excavation
entirely on subway tunnels in the central area of the
Peoples Square. IS-140.
Osborne, N., Ng, C. and Cheah, C. The benefits of hybrid
ground treatment in significantly reducing wall movement: a Singapore case history. IS-378.
Qiu, D., Zhou, K., Ding, Y., Liang, Q. and Yang, S. 2008. 3D
deformation monitoring of subway tunnel. IS-151.
Quick, H., Michael, J., Meissner, S. and Arslan, U. 2008.
Challenging urban tunnelling projects in soft soil conditions. IS-358.
Shao, Y. and Macari, E. 2008. Information feedback analysis in deep excavations. ASCE Int. Jou. Geom. Vol. 8, 1,
91103.
Wang, R., Cai,Y. and Liu, J. 2008. Supervision and protection
of Shanghai Mass Rapid Line 4 shield tunneling across the
adjacent operating metro line. IS-033.
Wong, K., Ng, N., Leung, L. and Chan, Y. 2008. Kowloon
Southern Link TBM crossing over MTR Tsuen Wan
Line tunnels in HKSAR. IS-370.
Xu, Q., Ma, X. and Ma, Z. 2008. Application of pile underpinning technology on shield machine crossing through
pile foundations of road bridge. IS-326.
Yoo, C., Kim, S. and Lee, Y. 2008. Characteristics of
tunneling-induced ground settlement in groundwater
drawdown environment. IS-329.
Zhao, H., Liu, X., Yuan, Y. and Chi, Y. 2008. Effect of longterm settlement on longitudinal mechanical performance
of tunnel in soft soil. IS-199.
129
Theme 1: Analysis and numerical modeling of

deep excavations
Optimization design of composite soil-nailing in loess excavation

G.M. Chang
Urban Construction and Environment Engineering Department, West Anhui University, LuAn, Anhui, P.R. China
School of Civil Engineering, ChangAn University, XiAn, Shanxi, P.R. China
ABSTRACT: The loess excavation has its unique characteristics compared with the others due to its structural property and collapsibility. In order to acquire the work mechanism and design methods of composite
soil-nailing in loess excavation, a reasonable finite element analysis model is selected. The optimization design
methods are introduced based on the results of finite element analysis conducted to determine the regularity
of deformation, the safety factor and the endogen force of the structure along with the change of design variable. Finally, the optimization design methods are validated contrasted with the data measured in an actual
project.
INTRODUCTION
2.1
Composite soil-nailing combined soil nails with

other forms of supporting measures has avoided the
soil-nailing technology from excessive dependence
on the soil and expanded its application field. Among
the different kinds of composite soil-nailing forms,
the anchor composite soil-nailing support method is
widely applied for its powerful location adaptability,
easy construction, low cost and reducing the pit deformation remarkably. However, its work mechanism and
design method, especially the Loess Pit anchor composite soil-nailing research, fall behind the project
practice by far. In the first instance, this paper aims at
studying nail design parameter selection in plain soilnailing on the premise of maintaining soil-nailing total
length, and then replacing a anchor for a soil nail to
research the parameter value of anchor composite soilnailing structure and optimization design under the
circumstances of maintaining plain soil-nailing design
parameters a more optimal value.
PARAMETER ANALYSIS
The overall stability and working performance of excavation supporting is closely related to the design
parameters. Understanding and grasping the relations
of the overall stability safety factor with the change of
these design variables, particularly this kind of sensitivity degree that variety, have special and important
meaning for guiding engineering practice.
Hypothesis
To simplify the calculations, we make the following

assumptions when carry on the numerical analysis to
the composite soil-nailing numerical analysis:
1 Composite soil-nailing problems are plane strain
problems;
2 Soil-nailing and assistance reinforcement materials
are elastic materials;
3 The soil is presumed as the elastic-plastic material.
2.2
Computation diagram and parameter

of material
2.2.1 Computation diagram

Engineering experience shows that the influence of
excavation width is about 3 to 4 times of the excavation depth, influence depth is about 2 to 3 times of
excavation depth. The case assumes that the excavation depth is 9.5 m, the total length of the finite element
model is 45 m, the total height is 25 m and the slope
gradient is 1:0.1 (Fig. 1).
2.2.2 Boundary conditions and loads
On the left and right boundary of the model, we set the
X-direction displacement to zero and allow theY direction deformation; the X and Y direction displacement
of the bottom boundary are zero; the top is a free surface. Initial stress field is gravity stress field; the value
of Gravitational Acceleration is 9.8 m/s2 . Since composite soil-nailing is usually constructed after precipitation, the impact of groundwater is not considered.
133
distance to ground/m
Figure 1. Finite element analysis model (Anchor at middle).
4
0
1
2
3
4
5
6
7
8
9
10
lateral deformation/mm
12
16
20
24
28
10
15
20
25
Figure 2. Relations of soil-nailing angle and pit

displacement.
2.2.3 Material parameters

This research was completed against the loess excavations and the soil parameters were provided from
a engineering investigation report in Xian City.
Because the soil distributes in certain scope are
uneven, it is discommodious to carry on the Numerical Calculation and take the soil strength average value
of each level. Soil nails and anchors take the form of
the commonly used procedure in Xian: Soil-nailing
110 mm diameter bored, steel bar 1 22; the anchor
hole diameter 150 mm, steel bar 2 18, surface 100
for C25 thick concrete, reinforced with distribution
steel bar network. Soil nail and anchor were made of
steel bars that wrapped with cement slurry composition. Slurry tightly wrapped the external part of steel
bars, and occluded with the soil in dogtooth. In order to
simulate the mechanical behavior of soil-nailing and
anchors correctly and simplify finite element analysis process, we regard the steel bar and the cement
paste body as a kind of compound material. Materials
geometric and mechanical parameters, as follows:
Soil: c = 30 kpa, = 18 , gravity = 18 KN/m3 ,
deformation modulus E0 = 1.8 107 Pa, Poissons
ratio = 0.3; Soil-nailing: diameter 0.11 m, sectional area is 0.0094985 m2 , moment of inertia
1.8324 106 m4 , equivalent modulus of elasticity
Eeq = 2 1010 Pa, Poissons ratio = 0.3;
Anchor: the sectional area of free segment is
5.0868 104 m2 , moment of inertia is 1.030077
108 m4 , elastic modulus Es = 2 1011 Pa, the sectional area of anchorage segment is 0.0176625
104 m2 , moment of inertia is 2.483789 105 m4 ,
equivalent elastic modulus Eeq = 2.03 1010 Pa.
Poissons ratio = 0.3; Surface (unit length) :
The sectional area is 0.1 m2 , moment of inertia
is 8.33333 105 m4 . Equivalent elastic modulus
Eeq = 2.1 1010 Pa, Poissons ratio = 0.3.
Contact surface friction: According to the research
of reference (Chen, 2000, Wang, 1997), the soil-nail
contact surface and the soil-anchor contact surface
friction value is 60 kPa.
Table 1.
Relations of soil-nailing angle and safety factor.
Soil-nailing
angle
Safety
factor
10
15
20
25
1.593
1.623
1.654
1.615
1.568
1.53
2.3 Soil-nailing support

2.3.1 The angle of soil-nailing
Concerning with the construction method, the angle
of soil-nailing has great influence on the pit
displacement, the safety factor and the surface subsidence. Taking the total length of soil-nailing is 40 m,
establishing five rows of soil-nailing, taking the soilnailing level and the vertical spacing takes 1.8 m, the
first row of soil-nailing depth of burying is 1.8 m.
Dividing five steps excavates, the first step of excavation depth is 2.3 m, and the other step of cutting
depth is 1.8 m each. Soil-nailing obliquities are calculated by inclination of 0 , 5 , 10 , 15 , 20 and 25
respectively.
Figure 2 and Table 1 show that the horizontal displacement is gradually increasing and changing at an
increasingly rapid pace as soil-nailing angle from 0
degrees to 25 changes gradually. When the pit design
requires strict control of the horizontal displacement,
they should use a smaller angle. Safety factor in
soil-nailing angle reduces 10 degrees at the largest
and declines rapidly with the angle increases after
10 degrees.
On the other hand, Soil-nailing angle is related
with construction methods and soil-nailing construction usually adopt the self grouting methods, in the
hope of soil-nailing has more inclination to make the
cement grout fill soil-nailing holes under the weight
easily.
134
5
0
10
15
20
25
Figure 3. Long at upper row and short at lower row.
2
3
4
5
6
7
long-short
mid long
short-long
8
9
10
Figure 6. Relations of soil-nailing layout and displacement.

Table 2. Relations of scheme of soil-nailing layout and
safety factor.
Scheme of layout
Long-short
Short-long
Mid-long
Safety factor
1.478
1.447
1.376
Figure 4. Long at middle row and short at upper and

lower row.
Selecting nail angle is 10 degrees then calculates

and analyzes on three different layouts, we get the
results Figure 6 and Table 2 below.
From Figure 6 and Table 2 we can find that there are
smallest displacement and largest safety factor when
using long-short scheme. On the contrary, there are
largest displacement and smallest safety factor when
using short-long scheme. As the same conclusion with
our forerunners, displacement and safety factor that
use short-long short scheme are between the other
two modes and we can see that when using plain
soil-nailing support a long-short scheme should be
adopted.
Figure 5. Short at upper row and long at lower row.
2.4 Anchored soil-nailing support
So after considering the pit displacement, the safety

factor and construction factors, the angle should be
about 1015 degrees.
2.3.2 Scheme of soil-nailing layout
Other researchers do more about the schemes of soilnailing layout (Hu & Song, 1997, Li & Zhang, 1999).
But their studies focus more on comparison between
long at upper row and short at lower row scheme (longshort scheme) and short at upper row and long at lower
row scheme (short-long scheme). But in practice, we
often use long at mid row and short at upper and lower
row scheme (mid-long scheme), especially when the
soil-nailing is used with anchor together (Figs. 35).
At present, the application of prestressed anchor in

composite soil-nailing design is often used empirically
and there has not a determinate calculative method
to set the anchor position, the length of anchorage,
prestressed value.
Based on previous studies, we chose the nails angle
of 10 degrees and long-short layout scheme to research
the anchored soil-nailing support.
2.4.1 Location of prestressed anchor
After replacing the 1st and 3rd rows and the fifth row
of soil nails with anchor, let us study the influence of
blot location on the pits level displacement and safety
factors. Diagrams shows in Figures 78, 1, and the
results are shown in Figure 11 and Table 2 below.
135
Figure 7. Anchor at upper row.
0
1
2
3
4
5
6
7
8
9
10
10
15
20
25
0 kN
100 kN
150 kN
200 kN
250 kN
Figure 10. Relations of prestressing and displacement.

Table 4.
lateral deformation /mm

10
15
20
0
1
2
3
4
5
6
7
8
9
10
25
soil
upper anchor
mid anchor
lower anchor
Figure 9. Relations of anchor locations and displacement.

Table 3.
Relations of anchor locations and safety factor.
Anchor locations
Upper
Middle
Lower
Safety factor
1.496
1.546
1.511
Prestressing
value/kN
100
150
200
250
Safety factor
1.546
1.597
1.617
1.624
1.628
plain soil-nailing, it can significantly reduce the

level of surface displacement through adding prestressed anchor, particularly the top-anchor scheme
and bottom-anchor scheme have the most obviously
effect on the surface of the horizontal displacement
control. Anchor locations also affect the safety factor. It has the biggest safety factor when anchor at the
central pit.
Therefore, to control the pit deformation, the angle
of anchor would favor the upper-anchor scheme; the
mid-anchor scheme is the most beneficial to improve
the safety factor. However, in the engineering practice, because excavations concentrate more and more
on urban areas and the anchor may into the pit slope
outside more distance, there may affect anchor construction for the adjacent buildings when used the top
anchor scheme.
Figure 8. Anchor at lower row.
Relations of prestressing and safety factor.
Figure 9 and Table 3 show that add prestressed

anchor into soil-nailing can significantly reduce the
maximum horizontal displacement, especially in and
near the anchor location. In addition, compared with
2.4.2 Level of prestressing value

Used prestressed anchor replace with the 3rd soilnail, taking the anchor free segment length is 10 m,
anchorage segment length is 8 m, prestressing value
is 0 kN, 150 kN, 200 kN and 250 kN respectively for
calculating. The results calculated from the horizontal displacement and the safety factors are shown in
Figure 10 and Table 4.
Figure 10 and Table 4 show that the impact of
prestressing value on the horizontal displacement
is greater. When the magnitude of prestressing is
100 kN, horizontal displacement decreases more than
plain soil-nailing from the maximum displacement
of 23.1 mm to 17.6 mm lower. If prestressing value
136
0
0
1
2
3
4
5
6
7
8
9
10
8
12
16
soil-nailing is a special kind of support, which between

plain soil-nailing and prestressed anchoring support. It has the advantages of both the plain soilnailing support and the prestressed anchor support
simultaneously.
20
3.1 Anchor and soil-nailing working together

6m
8m
10 m
12 m
14 m
Figure 11. Relations of anchorage segment length and

displacement.
Table 5.
factor.
Relations of anchorage segment length and safety
Anchorage
length/m
Safety
factor
10
12
14
1.586
1.617
1.629
1.637
1.641
increases to 150 kN, the maximum horizontal displacement will decrease to 15.3 mm. If the prestressing
value is more than 150 kN, such as 200 kN, 250 kN,
horizontal displacement will continue to decline, but
the reduction range is insignificant. As the value of
the prestressing increases, the Pit safety factor will
gradually increase in a limited extent, which means
the value of prestressing has no significant impact on
safety factor.
2.4.3 The length of anchorage segment
Selecting a middle-anchor scheme, the prestressing
value is 150 kN, taking the length of anchorage for
6 m and 8 m, 10 m, 12 m, 14 m to calculate, the results
are shown in Figure 11 and Table 5.
Figure 11 and Table 5 show that the horizontal displacement of Pit gradually diminishes as the anchorage
length increases, but the reduction is modest. Pit safety
factor would increase as the length of anchor increased
either, but not markedly.
It is at the initiative stressful condition as anchor support construction completed because of the existence
of prestressing. As a result of the anchor prestressing reaction, the soil is caused to be at the pressed
condition, reduced soil lateral deformation. On the
other hand, anchor is wrapped in the cement paste,
and adhibited with cement paste. Because of the holes,
pores and crannies existed in soil; the cement and
the soil assume the zigzag linking. After anchor tension deformation, there will have a shear stress due
to elastic deformation and retraction in the anchorsoil interface, which direction on the soil deformation
under shear stress is the contrary. It reduces soil internal tensile stress, and will also limit the deformation of
soil. The axial force of soil-nailing is related with the
deformation of the earth. Because stress reduces the
soil deformation, soil-nailing internal force is reduced
more remarkably than plain soil-nailing .The closer
the anchor approaches the soil nails, the more the
axial force decreases. Therefore, the role of restrictions
pit deformation is the base of anchor and soil-nailing
working together.
3.2 Anchor contribution to resistance moment
When the Pit Slope in the event of damage, the slip
surface have too much plastic deformation to make
slide mass along for destruction under sliding. Generally, the sliding moment entirely depends on the depth
of excavation and the soil gravity. For a certain pit, the
soil depth and its weight usually are constant and its
sliding moment can be seen as a constant. Meanwhile
resistance moment is provided by the undisturbed soil,
shear strength, soil nails and anchor.
The contribution of soil nails performance lies in
three main aspects: soil-nailing presence gives the
slip surface place to the post-transfer slip, improves
the sliding area and increases the friction of slip surface. Uplift role of the soil-nailing that outside the slip
surface, and the bending resistance role of soil-nailing.
The contribution of anchor main features (Chang,
2007):
3 ANCHOR AND SOIL-NAILING WORKING

TOGETHER MECHANISM
Plain soil-nailing support is a passive support system and soil-nailing would have a role only when the
soil generates sufficient deformation. Anchor belongs
to the initiative support system and through prestress to control soil deformation.Anchored composite
137
1 The anchors anchorage is long in general and

extends into the steady soil mass in central slip away
from the excavation surface to provide a stronger
uplift capacity.
2 The anchors prestressing makes slide and stability
soil mass tightly squeezed each other to improve
the friction resistance to sliding and increase the
resistance moment.
3 When the slip surface crosses the anchorage

segment, the anchor resistance to bending has
some contribution, but the contribution is weak in
general.
3.3 The impact of prestressing to soil-nailing

axis force
There are many studies about the impact of prestressing to soil-nailing axis force and the conclusions are
the same. Generally, the prestressing will significantly
reduce soil-nailing forces and the greater the value of
prestressing, the smaller the soil-nailing internal force.
And moreover, the closer the anchor, the greater the
internal force reduction (Zhang & Liu, 2002, Zhen
et al. 2005).
DESIGN OPTIMIZATION
1. Soil-nail design. Recommending taking the soilnailings long-short layout and it is advisable to
select 1015 degrees.
2. Loess has strong structure, which should be used
as much as possible.
3. Anchor should be installed close to the central vertical part of the pit, which can achieve a higher safety
factor and restrict the pit deformation.
4. It does not provide a greater safety factor and better
control deformation even the anchor length is too
long. So it had better be about 812 m.
5. Since the anchor prestressed reaction limits soil
lateral deformation, reduces the lateral displacement of composite soil-nailing retaining and axial
force of the-soil-nailing near the anchor. On the
premise of meeting deformation control request,
we can shorten the length that several soil-nailing
top of the pit appropriately in long-short scheme,
long-short-long scheme while forming to reduce
project cost. However, in order not to reduce safety
factor, it is not recommended to shorten the length
of soil nails in the lower side.
6. The fixing on the prestressed value must be according to the soil shear strength values, it will be about
100 kN to 200 kN as well. Due to the prestressed
value has no obvious influence on project cost, we
can incline safety to choose the values greater. But
too much prestressed has no significant impact on
pit retaining performance.
5
5.1
ENGINEERING ANALYSIS
Project overview
Table 6. Site
characteristics.
layer
structure
and
geotechnical
Soil
class
Thickness/
(m)
Unit
weight
(kN/m3)
Cohesion/
kPa
Angle of
internal
friction/
Miscellaneous
LoessQ32EOL
LoessQ32EO L
Ancient soilQ3lal
LoessQ3al+PL
0.90
6.60
1.70
3.40
8.00
18.20
16.20
18.00
18.90
19.50
25.00
28.00
26.80
32.20
20.00
18.00
18.00
18.10
17.60
18.00
by north from the pit 6.3 m is a seven-storey masonry

structure residential buildings, the lime soil foundation depth is about 3 m.On the south-east there is a
18-storey high-rise building with one-storey basement
which depth is 6 m, reinforced concrete pile foundations are 36 m long, from Pit 9.10 m, and the adjacent
side of the project pit used soil-nailing in the construction; north of the seven-storey residential building
masonry structure, and roughly parallel to pit edges,
buildings length is 42 m and the width 13 m, the nearest to excavation is 5.3 m; the west side of the South
and an adjacent hotel podium which is a two storeys
building with an underground layer from the edge of
pit 4 m, framework and infrastructure end elevation
is 7.13 m; the south side is close to a main road,
there are water and gas pipelines under the sidewalks.
5.2 Engineering geological conditions
According to geotechnical engineering investigation
report that the project site geomorphic units belong to
the Loess beam-swamp landscape. Proposed site layer
structure and geotechnical characteristics are in range
of 30.0 m deep in Table 5.
5.3 Retaining design
The design of anchored composite soil-nailing adopts
the methods proposed in part 4 of this paper, which is
the excavation depth was 11.0 m and the slope was
1:0.1. The basic design parameters were shown in
Table 6. There are six layers of soil nails and the layout
is cinquefoil. We also set a prestressed anchor at the
depth of 6.0 m to reduce the lateral displacement and
ensure the high-rise buildings in safety and stability.
Using two 18 mm diameter grade 60 bars in anchor
(10 m free, 8 m anchorage); Using one 22 mm diameter grade 60 bar in soil-nailing with 1.5 m spacing and
inclination of 15 degrees, prestressing value is 150 kN.
(Fig. 12).
5.4 Monitoring results
A project in Xian , the excavation depth is 11.0 m, both

the Pits length and width are about 100 m. In the east
This project was constructed since March 20, 2006 and

lasted 75 days. Monitoring results showed the greatest
138
Anchor length should not be too long and prestressing

also should not be too large. Practical project proved
that this optimization method in loess pit is applicable.
ACKNOWLEDGMENT
This paper was supported by Youth Foundation of
Anhui Education Committee (No. 2007jql181), and
supported by Youth Foundation of West Anhui University (No. wxzq2006018).
Figure 12. Composite soil-nailing support diagram.
REFERENCES
lateral displacement occurred in the east central pit and

the largest displacement was 16 mm. It was no excessive lateral deformation and earth surface subsidence
and surrounding buildings were no greater settlement.
6
CONCLUSIONS
In this paper, the author analyzed the parameter sensitivity of composite soil-nailing in loess excavation
using the finite element method. The optimization
design methods are introduced based on the results of
finite element analysis. The writer believes that anchor
should be installed in the pit in the central vertical,
which can achieve a higher safety factor and restrict
the pit deformation obviously.
We can shorten the length that several soil-nailing
top of the pit appropriately in long-short scheme, midlong scheme while forming to reduce project cost.
Chang, G.M. 2007. Study on the Application of Composite soil-nailing in Loess Excavations. XiAn: ChangAn
University
Chen, Z.Y. 2000. The application of soil-nailing in excavations. Beijing: China Architecture & Building Press
Hu, K.G. & Song, Q.G. 1997. Nonlinear analysis of action
mechanism of soil-nailing wall . Industrial Construction
27(11): 1013
Li, S.H. & Zhang, Y.J. 1999. Numerical simulations by 2D
FEM in process of excavation and supporting of deep
foundation ditch. Chinese Journal of Rock Mechanics and
Engineering 18(3):342345
Wang, B.Y. 1997. Design of soil-nailing. Geotechnical Engineering Technique (4): 3041
Zhang, F., Liu, Z.C. & Chen, G.G. 2002. The mechanical
working mechanism research on the united supporting of
prestressed soil anchor and soil-nailing. Rock and Soil
Mechanics 23(3): 292296
Zheng, Z.H., et al. 2005. In-situ testing study on retaining
miscellaneous fill slope by using compound soil-nailing.
Chinese Journal of Rock Mechanics and Engineering
24(5): 898904
139
Three-dimensional finite element analysis of diaphragm walls for

top-down construction
J. Hsi, H. Zhang & T. Kokubun
SMEC Australia Pty Ltd, Sydney, NSW, Australia
ABSTRACT: The Tugun Bypass Tunnel in Gold Coast, Australia was constructed using diaphragm walls with
the top-down cut-and-cover method to allow simultaneous construction of an airport runway extension above
the tunnel, whilst excavation of the tunnel continued underneath. The tunnel was built in an environment of
high groundwater table and deep deposits of alluvial and estuarine soils with the toes of the walls founded
in soil deposits. There was a potential risk for differential settlements between the diaphragm wall panels,
caused by the runway fill placed over the tunnel roof during excavation. Three-dimensional numerical modelling
was undertaken to predict the differential settlements of the tunnel with considerations of varying subsurface
profile, staged excavation and dewatering, non-uniform loading and complex soil-structure interaction. Field
instrumentation and monitoring was implemented to confirm numerical predictions.
INTRODUCTION
The 7 km long Tugun Bypass forms part of the

Pacific Highway, and connects south-east Queensland
to northern New South Wales, Australia. One of the
key features of the project was a tunnel of about 334 m
in length (Ch5588 to Ch5922.4), constructed below
the proposed runway extension of the Gold Coast Airport. Figure 1 presents the project route plan showing
the locality of the project.
As the tunnel was to be constructed in the proximity

of the airport runway, there was a strict height restriction for the construction activities. Low headroom
plant and equipment were chosen to construct the
diaphragm walls for the cut and cover tunnel. As the
construction of the runway extension coincided with
the tunnel construction, the top-down construction
method was adopted.
The subsurface of the tunnel site comprised mainly
alluvial and estuarine soils up to depths of about
35 m underlain by weathered rock of Neranleigh Fernvale formation. To minimize construction costs, the
diaphragm walls were founded in soil deposits which
were subjected to settlement under the loading from
the runway extension.
Excessive differential settlement of the diaphragm
walls could overstress the tunnel structure and affect
the tunnel serviceability. Detailed numerical modelling was carried out using the finite element package
PLAXIS 3D Foundation (Version 1.6) where the spatial subsurface variation and non-uniform loading
patterns could be taken into consideration.
Instrumentation and monitoring were undertaken
to demonstrate the field performance, which was then
compared with the numerical predictions.
2
Figure 1. Project route plan.
SITE GEOLOGY
The tunnel was situated in a flood plain which was

subjected to periodical flooding. The geology of the
site comprised Neranleigh Fernvale Beds overlain
by Cenozoic estuarine and coastal deposits. These
141
Figure 2. Site investigation plan.
deposits were up to 35 m in thickness, comprising river

gravels, sands and clays, and flood plain and tidal delta
muds and silts. At the tunnel location, the subsurface
horizons consisted of dune sands, Coffee Rock
(local term given to cemented silty sands), estuarine
interbedded clays and sands, and residual soils derived
from the weathered bedrock. Groundwater was slightly
saline due to the close proximity to the ocean. The
water table was influenced by both tidal movements
and rainfall events recharging Cobaki Broadwater.
Due to low-lying ground surfaces, potentials existed
for acid sulphate soils.
3
Figure 3. Subsurface profile.

Table 1. Typical soil profile and key geotechnical
parameters.
GEOTECHNICAL MODEL
As the subsurface conditions varied spatially along the

length and width of the tunnel, extensive site investigations using boreholes (BH) and piezocones (CPTU)
were undertaken at the wall and barrette locations.
Within the footprint of the runway extension, the investigations were done at a spacing of approximately 20 m
intervals. The plan of the site investigation is shown in
Figure 2.
The geotechnical model of the site included subsurface stratigraphy and geotechnical parameters.
The subsurface was divided into discrete soil units,
classified according to material type and consistency
or density and is summarized as follows (top down):
Top soil thin skinned (<1 m) comprising peaty
sandy organic topsoil, having loose consistency.
The ground surface was marshy and generally
untrafficable;
Dune sands a sequence of generally loose to very
loose sands of up to about 8 to 10 m in thickness,
fine to medium grained sands;
Coffee Rock (CR) a sequence of medium
dense to very dense cemented silty sands of about
7 to 10 m in thickness with occasional loose
consistency;
Estuarine a sequence of about 15 m thickness
comprising shell fragments, sand and silty sand,
clay and sandy clay, silt and clayey silt, clayey silty
sand and gravels. Relative density varied from very
loose to dense, and consistency varied from firm to
very stiff;
RL
(m)
0.5
4.0
11.2
13.5
17.5
21.1
23.0
28.6
30.8
Soil

(deg)
k
(m/day)
Eref
50 &
Eref
oed
(MPa)
Eref
ur
(MPa)
Sand (VL)
CR (MD)
CR (D)
Sand (L)
Clay (St)
Sand (L)
Clay (F)
Clay (VSt)
Bedrock
30
32
34
32
28
32
24
29
1.0
0.1
0.1
1.0
1 104
1.0
1 104
1 104
10
50
80
30
10
30
7
25
30
150
240
90
30
90
21
75
Note: RL (reduced level) is at top of each layer; VL is very

loose; L is lose; MD is medium dense; D is dense; F is firm;
St is stiff; VSt is very stiff; is drained friction angle; k is
permeability; Eref
50 is secant Youngs modulus at a reference
pressure of 100 kPa; Eref
oed is tangent Youngs modulus for primary odometer loading at a reference pressure of 100 kPa;
and Eref
ur is unloading/reloading Youngs modulus at a reference pressure of 100 kPa. Refer to PLAXIS manual for
Hardening Soil (HS) model.
Residual soil comprising clay and silty clay

with some sands, and with residual fragments of
extremely weathered and extremely low strength
interbedded argillite and greywacke of the Neranleigh Fernvale Beds. The thickness ranged between
about 1 m and 6 m;
Bedrock comprising extremely weathered to
moderately weathered and extremely low to low
strength interbedded argillite and greywacke, having an irregular contact with the overlain residual
material at a depth of approximately 30 to 35 m.
The subsurface profile based on the boreholes
along the centre line of the tunnel is presented in
Figure 3. The geotechnical parameters for each of the
units were determined from interpretation of the field
and laboratory test results, and based on local experience. The typical soil profile and key geotechnical
parameters assumed are shown in Table 1. The ground
142
surface level was approximately at RL 0.5 m and the

groundwater table was at the surface.
4
ISSUES AND CONSTRAINTS
Construction of a tunnel in soft ground at shallow

depths is conventionally undertaken using the cut and
cover method. However, to allow for construction
of the runway extension that occurred concurrently
with the tunnel excavation, the top-down construction
method had to be adopted. Diaphragm walls and cast
in situ tunnel roof slabs had been chosen to facilitate
the construction requirements and time constraints.
Figure 2 shows the footprint of the runway extension
oblique to the tunnel alignment.
Following the handover of ground surface, up to
23 m of fill for the airport runway extension was
placed above the tunnel roof. Loads acting over the
entire width of the roof slabs were transferred directly
to the diaphragm walls and the barrettes. The site
investigations revealed presence of estuarine deposits
consisting of loose materials below the toe of the
walls. Therefore, there was a potential for the tunnel
to settle during excavation. One of the critical issues
was the differential settlements between the walls and
the central barrettes, and along the walls. These differential settlements could potentially induce significant
stresses in the roof structures and in the walls.
Other issues in relation to the tunnel construction
are listed below:
Obstacle Limitation Surface (OLS) applied at
both ends of the runway to provide safe airspace for
approaching aircrafts. This required all construction activities to be undertaken within a headroom
of as low as 8 m. Use of cranes or heavy-lifting
equipment was only allowed outside the airport
operating hours;
High groundwater level due to its close proximity
to the sea and Cobaki Broadwater. The groundwater
was practically at the ground surface level. Reliable
dewatering system was essential during excavation;
Environmental requirements strict environmental controls were enforced such that drawdown of
the groundwater table outside the diaphragm walls
was insignificant. Also, all acidic sulphate soils
excavated from the tunnel had to be dried and
neutralized with lime prior to placement as fill in
embankments.
5
CONSTRUCTION METHOD
Suitable construction methods were chosen to address

the issues and constraints mentioned above. In order to
adhere to the OLS requirements, special low headroom
hydraulic grab (Leibherr HS852HD) and 2.8 m wide
trench cutter (CBC25) were used.The guide walls were
Figure 4. Diaphragm wall construction sequence (courtesy

of Bauer/Piling Contractors).
built first followed by construction of the 6 m wide

primary panels (Steps 1 to 5 of Figure 4) and 2.8 m
wide secondary panels (Steps 6 to 8). The open trench
was supported by mixture of bentonite slurry, when the
cutter undertook full excavation (Steps 2 to 3 and 6). A
steel reinforcement cage was lowered when the panel
was excavated to full depth (Steps 4 and 7). Concreting
of the panels was then achieved by the tremie method
(Steps 5 and 8). Figure 4 presents the construction
sequence of the diaphragm wall.
Following completion of the diaphragm walls
and barrettes, dewatering and excavation commenced
inside the walls. Excavation was initially undertaken
to depths of up to about RL 2 m to allow for construction of the roof slab. Water-tight membrane was
installed as part of the water-proofing system. When
the roof slab was completed, it was backfilled and the
site was cleared for handover to the Gold Coast Airport. These activities commenced in April 2006 after
environmental approvals were granted, and were completed by November 2006 which was the scheduled
date of handover of the site surface. Excavation below
the runway extension continued through to January
2007, and the remaining construction of the tunnel
continued.
6
STRUCTURAL DETAILS
The tunnel structure consisted of diaphragm walls

and barrettes located at the centre of the tunnel. The
diaphragm walls were 1 m in thickness, and extended
from the Northern Portal (Ch5588) to the Southern Portal (Ch5922.4). The walls were installed to
the depth of RL 17 m, from the top of the guide
wall at RL 2 m. The internal width between the
diaphragm walls ranged from about 25.7 m at the
northern portal to 28 m at the southern portal. Barrettes were 0.8 m thick and 2.8 m wide with a clear
spacing of 2.8 m throughout the central axis of the tunnel, extending to RL 17 m in depth. These structures
had a 100 year design life, using N-grade reinforcing
143
8.2 Three-dimensional numerical modelling
Figure 5. Typical tunnel cross section.
steels and 50 MPa high strength concrete. There were

no mechanical joints at the interface of the primary
and secondary panels in the longitudinal direction.
However, the barrettes and the diaphragm walls were
rigidly connected to the 1 m thick roof slab. There were
three jet fan niches where the roof slab was slightly elevated. The base slab was also 1 m thick with a founding
level ranging from RL 5.5 m to RL 9.5 m. Figure 5
shows the typical cross section of the tunnel.
7
DESIGN CONSIDERATIONS
Geotechnical design of the tunnel was required to

satisfy the following three key issues:
Excavation support during construction the
diaphragm wall structures were designed to ensure
stability of the excavation. Issues included structural design of the walls, base heave, hydraulic
uplift, piping, and liquefaction;
Long term stability of the tunnel buoyancy of the
tunnel when the groundwater table was close to the
surface;
Serviceability assessment due to settlement of the
tunnel during construction the tunnel was subjected to loading from airport runway fill which
resulted in settlements. The influences of differential settlements on structural capacity were
assessed.
8
NUMERICAL MODELLING
8.1 Two-dimensional numerical modelling

Design of the tunnel was initially undertaken using
the finite element software PLAXIS (Version 8.4) at
selected sections. This numerical package was used
to analyze two-dimensional plane-strain problems
involving complex soil-structure interaction for the
design of the structural members. Structural beam
elements were used to simulate the diaphragm walls.
Global factor of safety during each of the construction
stages was calculated based on the c - reduction
method to ensure the minimum FoS was achieved. The
software allowed modelling of construction sequence,
changing groundwater levels, and varying subsurface
conditions across the width of the tunnel.
A three-dimensional numerical modelling package,

PLAXIS 3D Foundation (Version 1.6), was employed
to predict the settlements of the tunnel caused by runway fill loading and excavation. Due to the limitation
of the program, settlement analyses were undertaken
in sections, each of approximately 40 to 60 m in
length. The major advantages of the 3D modelling
were as follows:
Ability to model the physical dimensions of the wall
and barrette structures. This improved the accuracy
of settlement prediction, as it accounted for longitudinal stiffness of the tunnel which assisted in load
redistribution and toe resistance of the structures;
Ability to simulate 3D load distribution where the
runway fill was placed oblique to the longitudinal
axis of the tunnel;
Ability to model 3D subsurface profile based on
probe holes at discrete locations;
Ability to simulate dewatering within the tunnel
excavation area.
The Hardening Soil (HS) model was considered
most appropriate to simulate soil behaviour in an open
excavation. The HS model took into account unloading
and reloading behavior and irreversible plastic strains
of soil. The HS stiffness parameters were defined with
respect to a reference pressure of 100 kPa. The key
ref
ref
parameters included Eref
50 , Eoed , and Eur as shown in
Table 1. The published data indicate the ratio of Eref
oed
ref
ref
to Eref
50 is about 0.7 to 1.4 and the ratio of Eur to E50
ref
ref
varies from 2 to 4. The analysis adopted E50 = Eoed ,
ref
and Eref
ur = 3E50 .
Presented here is a 41.2 m long section of the tunnel between Ch5728.8 and Ch5770. This section of
the tunnel was of the deepest location of the tunnel, beneath the thickest layer of the runway fill, and
underlain by sloping bedrock level and changing clay
thickness. A jet fan niche of approximately 12 m long
also lied within the centre of this section which had also
been incorporated in the model. Within this chainage
range, there were seven boreholes. Due to the capacity
of the program, four representative boreholes, which
were evenly distributed spatially, were selected for the
analysis. The assumed subsurface profiles are shown
in Table 2.
8.3 Assumptions of analysis
The construction sequence was considered in the analysis to simulate the load transfer from the runway
fill to the diaphragm walls. The assumed construction
sequence is described below:
1. Application of loads exerted on the virgin ground
from the working platform built to RL 2 m (for
construction of the guide walls) and construction
load of 10 kPa;
144
Table 2.
Table 3. Assumed geometry during construction.
Subsurface profiles.
Borehole #1
Location LHS
Chainage 5730
#2
RHS
5737
#3
LHS
5757
Unit
RL at top of each layer
1
2
3
4
5
6
7
8
9
0.5
4.8
10.8
13.8
17.8
24.3
27.3
33.3
36.3
0.5
3.4
14.4
15.9
17.1
21.8
25.1
30.8
32.0
0.5
5.0
9.8
12.6
17.5
20.8
21.8
25.0
27.4
#4
Centre
5768
0.5
2.8
9.7
11.7
17.5
17.5
18.0
25.2
27.4
5728.8 5737.6
to
to
5737.6 5743.6
Soil type
(density/
consistency)
Sand (VL)
CR (MD)
CR (D)
Sand (L)
Clay (St)
Sand (L)
Clay (F)
Clay (VSt)
Bedrock
Note: LHS is left hand side of tunnel facing increasing

chainage direction; RHS is right hand side of tunnel; and
Centre is centre line of tunnel.
2. Installation of diaphragm walls and barrettes to RL

17 m;
3. Removal of the working platform, and application
of 10 kPa construction load on surface;
4. Dewatering and excavation to underside of the roof
slab;
5. Installation of the roof slab (and jet fan niche), and
backfill to existing ground surface;
6. Placement of runway fill to design heights (simulated as pressures) with 10 kPa live load above the
runway;
7. Staged dewatering and excavation within the
diaphragm walls to underside of the base slab;
8. Casting of the base slab and completion of the
tunnel structure;
9. Return of the groundwater table to the ground
surface and removal of 10 kPa surface loads.
Chainage range
Feature
5743.6 5755.2
to
to
5755.2 5761.2
5761.2
to
5770.0
RL (m)
Natural Ground
0.5
Level
Top of Roof Slab
0.8
Bottom of Roof
1.8
Slab
Initial Excavation 2.8
Initial Dewatering 3.8
Intermediate
6.0
Excavation
Intermediate
7.0
Dewatering
Top of Base Slab
8.4
Bottom of Base
9.4
Slab
Final Excavation
9.7
Final Dewatering 11.7
Toe of Diaphragm 17.0
Wall
0.5
0.5
0.5
0.5
0.25
1.25
+0.4
0.6
0.25
1.25
0.8
1.8
2.8
3.8
6.0
2.8
3.8
6.0
2.8
3.8
6.0
2.8
3.8
6.0
7.0
7.0
7.0
7.0
8.4
9.4
8.4
9.4
8.4
9.4
8.4
9.4
9.7
11.7
17.0
9.7
11.7
17.0
9.7
11.7
17.0
9.7
11.7
17.0
Figure 6. Deformed 3D finite element mesh.
The settlement assessment was undertaken at

stage 7, which was considered most critical with
maximum excavation under full runway loading.
The assumed levels within the modeled chainage
range are summarized in Table 3.
for uncertainties, the tunnel was designed for a maximum differential settlement of 25 mm. The structural
analysis showed that the longitudinal in-plane stiffness
of the tunnel would smooth out differential settlements
along the tunnel alignment, with the presence of the jet
fan niche and variability of the subsurface conditions.
8.4
Results of analysis
The deformed mesh of the 3D finite element analysis

under the full runway loading and at the final stage
of the excavation is shown in Figure 6. The predicted
settlement profiles at the top of the roof slab along the
diaphragm walls and barrettes prior to casting of the
base slab are presented in Figure 7.
The predicted settlement of the tunnel during excavation was about 45 mm on the LHS, 43 mm on the
RHS, and 35 mm along the central barrettes. The
maximum differential settlement was predicted to be
12 mm between the walls and the barrettes. To allow
FIELD PERFORMANCE
The performance of tunnel during construction was

assessed based on the field monitoring results. This
was a means to confirm that the structural integrity of
the diaphragm walls and barrettes were not adversely
affected by differential settlements. Three instrumentation arrays were set up at Ch5655, Ch5718, and
Ch5770 corresponding to locations of the runway fill
(see Figure 8).
Each array consisted of three settlement plates
placed above the LHS and RHS diaphragm walls
145
10
Distance along Centre Line (m)

15
20
25
30
Date
35
40
45
26/11/2006 16/12/2006 5/01/2007 25/01/2007 14/02/2007 6/03/2007 26/03/2007 15/04/2007
10
Settlement (mm)
Settlement (mm)
35
40
45
20
30
RHS Diaphragm Wall
CH5770
50
Predicted 43mm (RHS)

Predicted 45mm (LHS)
60
Figure 7. Predicted settlement profiles at top of roof.
2
4
LHS Diaphragm Wall

55
Predicted 35mm (Center)
Central Barrettes
CH5728.8
2
Measured at LHS
Measured at Centre
Masured at RHS
RL of Excavation
RL of Top of Fill
10
40
50
6
8
RL of Excavation and Top of Fill (m)
30
10
Figure 10. Settlement monitoring results at Ch5718.
the result of analysis between Ch5728.8 and Ch5770.

Monitoring commenced at the beginning of November 2006. Excavation of the tunnel commenced in mid
December 2006 from the Northern Portal at Ch5588.
The excavation process reached Ch5718 in early January. Placement of runway fill above CH5718 followed
in mid January, which had resulted in visible settlements of the tunnel. The settlements appeared to have
ceased after the excavation reached final depth in mid
February. The monitoring data showed that the field
performance of the tunnel was consistent with the predictions obtained from the PLAXIS 3D Foundation
modelling. Maximum differential settlements between
the barrettes and the diaphragm walls were less than
25 mm at all stages of construction.
Figure 8. Plan of instrumentation arrays.
10
Figure 9. Typical instrumentation section.
and the central barrettes (see Figure 9). These were

installed prior to runway fill placement and excavation of the tunnel in order to capture all construction
induced movements. In addition to the settlement
plates, survey targets were also installed at inner walls
to the tunnel to record tunnel movement during excavation. This information had to be calibrated against
the settlement plate measurements as the initial tunnel
movement record was not available.
Figure 10 shows a summary of construction
activities, recorded settlements, and the predicted
settlements at diaphragm wall and barrette locations
at Ch5718. The settlement prediction adopted here is
CONCLUSIONS
The Tugun Bypass tunnel had to be constructed under

many strict constraints in a challenging geotechnical environment. Extensive site investigations were
undertaken to better characterize ground conditions
and reduce risks of geotechnical uncertainties. The
top-down construction method was adopted to allow
extension of the airport runway to occur simultaneously during tunnel construction. The additional loads
from the runway fill induced settlements of the tunnel
during construction. Settlement analysis of the tunnel
using 3D numerical modelling techniques had been
undertaken. The differential settlements of the tunnel
were successfully predicted. The performance of the
tunnel was monitored during construction and the field
measurements were consistent with the numerical
predictions.
REFERENCES
PLAXIS 2D, Version 8.4, PLAXIS BV Netherlands, 2006.
PLAXIS 3D Foundation, Version 1.6, PLAXIS BV
Netherlands, 2006.
146
Numerical evaluation of dewatering effect on deep excavation in soft clay

L. Li
Tianjin Institute of Ubran Construction, Tianjin, P.R. China
Tianjin Key Laboratory of Soft Soil Characteristics and Engineering Environment, Tianjin, P.R. China
M. Yang
Key Laboratory of Geotechnical and Underground Engineering of Ministry of Education, Tongji University,
Shanghai, P.R. China
Tongji University, Shanghai city, P.R. China
ABSTRACT: This paper describes the application of FLAC3D analysis for modeling a top-down construction
of a four-story (33.7 m depth), underground transformer substation in downtown of Shanghai city. There is
unconfined aquifer and confined aquifer on the site of this project and drainage by desiccation in the foundation
pit is adopted. The effective stress methods of analysis incorporate excavation and dewatering of the foundation
pit for real-time simulation of construction activities. The results not considering dewatering are compared with
the result considering dewatering, including wall deflections, basal heave, surface settlement. The analysis of
considering leakage of the wall and leakage between confined aquifer is provided also. The analysis shows
that although the difference is small in soft clay due to the low permeability of the soil, dewatering enhance
the deformation of the foundation pit and the foundation pit is inclined to be not security if dewatering is not
considered, the effect of leakage of the wall can be obvious on the surface settlement.
INTRODUCTION
For deep excavation in congested urban environments,

designers are particularly interested in making reliable prediction of the magnitudes of movements in
the surrounding soil (Peck 1969, Clough et al. 1989,
ORourke 1981) and then estimating the effects of
these movements on adjacent structures and facilities
(Burland & Wroth 1974, Boscardin & Cording 1989).
In principle these prediction can be achieved using
powerful numerical methods such as finite element
analyses, but there is difficulties in achieving reliable analytical predictions of soil deformations which
can be attributed to a variety of factors including
dewatering.
Dewatering is necessary in the excavation under the
ground water level in soft clay, which provide a dry
environment for the excavation and is benefit for the
slope stability and reducing the harm induced by the
groundwater, but dewatering have important effect on
the behaviour of the foundation pit and the surrounding soils, seepage induced by dewatering from outside
to the inside of the foundation pit have effect on the
stability and the deformation of the foundation pit and
surface settlement for the vertical consolidation by the
underground-water drawdown out of the foundation
pit, so effective stress methods of analysis distinct
from vast majority of analyses relies on total stress

methods of analysis and real-time simulation of coupling between ground-water flow (pore pressure)and
soil deformations is adopted in this numerical analysis considering dewatering. In this paper, considering
dewatering means considering the difference between
in and out of the foundation pit in the simulating, and
not considering dewatering means not considering the
difference.
This paper describes the application of FLAC3D
program for predicting soil deformations and ground
water flow associated with the top-down construction
of a columniform underground transformer substation
in downtown of Shanghai city. The model incorporates
a number of advanced features of analysis including:
(1) A FLAC3D model which can consider interaction
between support system of the foundation pit and the
surrounding soil; (2) Dewatering and excavation are all
considered in the numerical analysis at the same time;
(3) The drawdown of the ground water and the surface settlement due to dewatering in the foundation pit;
(4) The surface settlement induced by the leakage of
the diaphragm wall. Site stratigraphy, material properties, and initial ground-water condition are all selected
using information provided prior to construction, the
simulation of the construction sequence is based on
the scheme.
147
2.0m
3.2m
10.5m
16.5m
21.3m
26.5m
31.6m
5
6
12
37.5m
45.3m
1
2
3
4
11
Roof slab
Temporary bracing
Floor slab
Floor slab
Temporary bracing
Floor slab
Temporary bracing
Floor slab
7 1
7
Table 1.
2.0m
7.0m
11.5m
16.5m
22.0m
26.5m
31.0m
33.7m
60.3m
73.4m
77.4m
82.3m
100.0m
Input parameters in Mohr-coulomb model.

E
MPa
Stratum
C
kPa
( )
D
( )
1
2
1
2
1
2
3
1
2
1.2
9.3
8.9
8.2
11.4
14.9
26.8
44.9
80.2
16.4
28.3
72.2
99.2
133.4
0.36
0.36
0.37
0.39
0.38
0.37
0.34
0.32
0.31
0.34
0.33
0.33
0.31
0.29
0.32
25.9
5.1
8.0
13.4
27.1
42.7
5.0
0.0
21.2
16.9
22.1
0.0
0.0
22.5
17.4
21.2
19.7
15.7
15.8
13.7
31.2
33.0
23.1
24.1
20.3
35.0
37.0
0
0
0
0
0
0
0
0
0
0
0
0
5
7
Table 2.
Porosity and coefficient of permeability.

Coefficient of
permeability
Figure 1. Soil profile and the location of the wall and

bracing.
2
2.1
GENERAL SITUATION OF THE PROJECT

Project description
The columniform underground transformer substation

of a four floor and 33.7 m high underground structure
occupies a plan area of 13000 square meter (interior
diameter is 130 m) in the downtown of Shanghai city
and is bounded by buildings and viaduct and lots of
underground pipeline.
The underground transformer substation design
uses a cast in situ, reinforced concrete, diaphragm
wall (1.2 m thick) extending down into the elevation
57.5 m (With respect to the Shanghai City Base
datum), the circular wall is braced internally by the
floor slabs and 3 temporary annular bracing (Figure 1),
which are in turn supported by the interior columns
(steel and reinforced concrete & angel iron lattice)
founded on the bearing pile (bored filling pile, the pile
tip at depths 80 90 m below ground level. Both
the diaphragm wall and interior columns are installed
prior to excavation using slurry trench methods. The
roof and four floor levels are cast in sequence from the
top-down by excavating the soil from beneath the most
recently constructed slab. During excavation, dewatering is accomplished using drainage by desiccation.
2.2
Engineering geological and groundwater

conditions
Table 1 shows an averaged profile of subsurface

stratigraphy interpreted from borings conducted at the
site. It should be noted that the borings logs actually
Stratum Soil name
Vertical
Porosity m/s
Horizontal
m/s
11
12
1
1
2
1
2
3
1
2
0.56
0.49
0.57
0.58
0.52
0.51
0.43
0.46
0.44
0.51
0.50
0.47
0.37
0.35
5.5 107
3.5 106
8.1 108
4.1 108
3.2 108
4.1 108
2.6 105
3.8 105
4.0 106
2.8 106
3.0 104
3.0 104
3.0 104
Artifical soil
Silty clay
Mucky silt
Mucky silt
Clay
Silty clay
Silty clay
Silty silt
Silt sand
Silt clay
Silt sand
Silt sand
Medium sand
Coarse sand
2.5 109
1.7 108
7.2 109
3.9 109
4.0 108
5.8 109
2.6 106
5.4 106
8.2 109
6.2 108
1.7 106
5.7 106
7.9 106
show significant variations in the thickness of the individual strata across the site, The assumption of an
average profile is consistent with the limited area and
uncertainties in engineering properties of individual
strata. The material parameter of the layers is in table 1.
There is unconfined aquifer and confined aquifer
in the site. The ground water level of the unconfined
aquifer is 12 m under the ground and the 6th layer
is relative impervious layer. The confined aquifer is
divided into the 1st confined aquifer and the 2nd confined aquifer by the 1 2 layer. The 1st confined
aquifer lies in the 1 and 2 layer and the 2nd confined
aquifer lie in the 3 and layer, there may be some
relationship between the 1st and 2nd confined aquifer.
The porosity and the coefficient of permeability of the
layers is in table 2.
148
2.3
FLAC3D model description
The three dimensional numerical analysis program

FLAC3D is developed by Itasca Consulting Group,
Inc. The groundwater flow model may be coupled to
the stress model. The finite element model extends
far beneath the excavation (to 100 m depth) and laterally a distance of 200 m beyond the perimeter wall
where soil displacements, due to the simulation of
underground transformer substation construction, are
negligible. Constitutive modeling of soil behavior and
selection of input parameters represent a major source
of uncertainty in finite element analysis. The soil constitutive model use Mohr-coulomb failure criterion
and in order to model realistically the depth variations
in properties the elastic shear and bulk modulus are
assumed to be proportional to the mean effective confining stress. The unloading modulus of the soil in the
foundation pit is trinal loadings and the modulus of the
soil under the ultimate base of excavation adopt mixed
modulus because of tension pile. The parameters of
the soil are shown in table 1.
The soil and diaphragm wall adopt 8 nodes solid
element and the permanent floor slab adopted shell
element and the temporary annular bracing adopted
beam element, and interface is adopted in the joint
of the diaphragm and soil. Elastic model is used
for the diaphragm wall and the youngs modulus is
2.3 104 MPa, and the Poissons ratio is 0.167. Elastic
model is also used for floor slab and temporary annular
bracing and the parameter of them is consistent with
the scheme. The FLAC3D model of this project is in
Figure 2.
Boundaries condition is summarized as follows: (1)
The underside displacement of the model is zero, the
horizontal displacement of the side of the model is
zero, the upper surface is free; (2) The underside of the
model is impervious boundary, the side of the model is
pervious boundary and the pore pressure is fixed, the
pore pressure of the upper surface is fixed as zero in
the formation of the initial stress field and hydrostatic
pressure and is free during dewatering and excavation.
Dewatering in the foundation pit is simulated by controlling the saturation and pore pressure at specific
locations in the element model.
2.4
Construction sequence
Based on the actual record of site activities and the

sequence of events occurs in the finite element simulation, this process is simulated by 17 stages. Each
stage in the analysis represents a distinct change
in either the geometry, boundary conditions or time
elapsed between events. The first stage is the formation
of the initial stress field and hydrostatic pressure
and after every dewatering and excavation (including
adding bracing) is a stage. The numerical simulation
assumes that the construction of the diaphragm wall
Figure 2. FLAC3D model of the project.
has no effect on the surrounding soil (i.e., the wall

is wished-in place)and does not consider the installation of load-bearing elements used to support the
internal column.
3
RESULT AND ANALYSIS
3.1 Analysis of the seepage field

Due to unconfined aquifer and confined aquifer existing in the site, there is difficulty in simulating them at
the same time. The confined aquifer is not taken into
149
Figure 4(c). Contour of the lateral deflection.
Figure 3. The neural pressure contour and flow vector.
excavation stage. The maximum lateral deflection of

the wall is 23.3 mm, the ratio of the maximum lateral
deflection to the end excavation depth is 0.07%, the
top-down and the high rigidity of the slab is the reason
for the small ratio. The location of the lateral deflection
is a little higher than the excavation face and falling
with the excavating and the location of the maximum
lateral deflection is at 28.1 m. Figure 4(b) is the contrast of the lateral deflection considering dewatering
and not in the last 3 excavation stage. The maximum
lateral deflection considering dewatering is 27.7 mm
and higher than the result not considering dewatering.
The difference mostly happen in the middle and underside wall and is increasing with the excavation depth
for the seepage force is concentrated in the middle and
underside wall and the hydraulic head is increasing
with the excavation depth. Figure 4(c) is the contour
of lateral displacement of the diaphragm wall.
Figure 4(a). Lateral deflection of wall.
3.3 Analysis of the basal heave formation
account when analyzing the seepage of the unconfined

aquifer, which have no influence on the phreatic line.
The neutral pressure contour and the flow vector when
excavating to 33.7 m is as Figure 3. It shows that the
drawdown of the phreatic water is small and the depression cone is not obvious due to the low permeability
of the soft clay.
The basal heave is including the elastic rebounding

and the local plastic failure and the deep-seated plastic
failure. The elastic rebound is because of the unloading, the local plastic failure is that the soil near the
wall yield, and the deep-seated plastic failure is that
the soil in the bottom is short of bearing power. When
the excavation depth is little, the elastic rebound is
the most, the maximum heave lies in the center, with
the excavation depth increasing the heave in the circumference preponderate over the heave in the center
because of the failure in the circumference, Figure 5(a)
is the curve of the basal heave, Figure 5(b) is the contour of the basal heave when excavating to 33.7 m.
Figure 5(c) is the contrast of the basal heave result considering dewatering and not in the last two excavation
stages, it shows the basal heave considering dewatering is high than the not and the difference is increasing
with the excavation depth .
3.2 Analysis of the lateral deflection of wall
3.4 Analysis of the surface settlement
Figure 4 (a) is the computing result of the lateral deflections of the diaphragm wall along the depth at every
Figure 6(a) shows the surface settlement after the

every excavation stage. The surface settlement and
Figure 4(b). Contrast of the wall deflection.
150
Figure 5(a). The deformation of the basal heave.

Figure 6(a). Surface settlement at stages.
Figure 6(b). Contour of the surface settlement.

Figure 5(b). Contour of the basal heave.
Figure 5(c). Contrast of the deformation of the basal heave.
Figure 6(c). Contrast of the surface settlement.
the distance between the maximum surface settlement

and the foundation pit is increasing with the excavation depth. The maximum settlement is 15.7 mm and
the maximum distance is 44 m when excavating to
33.7 m depth. The diaphragm wall move up and raise
the soil near the foundation pit because of the unloading, which has a effect on the location of the maximum
surface settlement .
Figure 6(b) is the contour of the surface settlement.

Figure 6(c) shows the contrast of the surface settlement
considering dewatering and not. It shows that the magnitude and range of the surface settlement induced by
the dewatering is small for the reason of the low permeability of the soil and the small descent of groundwater
level induced by dewatering. The location of maximum surface settlement induced by dewatering is
closer to the foundation pit than by excavation and the
151
CONCLUSIONS
This paper introduce the FLAC3D model for simulating the top-down construction of an underground
transformer substation at Shanghai city. Results considering dewatering are compared with the result not
considering including wall deflection, basal heave,
and surface settlement, the effect of leakage is also
analyzed. The main conclusions of this study are as
follows:
Figure 7. Leakage in the 6th layer.
maximum surface settlement considering dewatering

is 15.9 mm.
3.5 Analysis of leakage
By the excavation experience in Shanghai city, the
leakage of diaphragm wall often occur when the depth
beyond 28 m and the probability increase with the
excavation depth. The main cause is that the bad joint
of the wall and the dimension error in construction
and the distortion of the diaphragm wall. The leakage
will induce the decline of the ground water level and
the additional surface settlement, so the effect of the
leakage have to be taken into account.
If the leakage occurs in the sixth layer, Figure. 7
shows the neutral pressure and the flow velocity vector, by the contour of the neutral pressure, only the
groundwater level near the foundation pit descend little, so the leakage in the 6th layer have little effect on
the settlement around the foundation pit because of the
low permeability and small discharge.
If the leakage occurs in the 1st confined aquifer
which of the water pressure is 5 6 m under the
ground , the water pressure of the 1st confined aquifer
descend clearly and the surface settlement is distinct
for the permeability of the 1st confined aquifer is high
and the discharge is much.
The 2nd confined aquifer can leak to interior of the
1st confined aquifer for the fall of the neutral pressure in the 1st confined aquifer due to dewatering and
excavating in 1 layer. The numerical analysis shows
leakage occurs, but the discharge is little, so the leakage have little effect on the settlement for permeability
of the of the 1 gray clay is low (8.21 109 m/s) and
the thickness of 1 layer reach to 15 m.
1. Dewatering by desiccation in the foundation pit

have effect on the behavior of the foundation pit, the
seepage could enhance the wall deflection and the
deformation of the basal heave, the drawdown of
the groundwater level outside of the foundation pit
could result in the vertical consolidation and add the
surface settlement. The foundation pit is inclined to
be not security if dewatering is not considered.
2. The wall deflection and basal heave due to the seepage is little because of the low permeability of the
soil in soft clay. The drawdown of the groundwater
level due to dewatering is little, which result in little
surface settlement for the same reason above.
3. The leakage of the wall have important effect on
the surface settlement, which is a problem needed
to solve.
REFERENCES
Andrew J. Whittle, Youssef M. A. Hashash & Robert V.
Whitman. 1993. Analysis of deep excavation in boston,
Journal of Geotechnical Engineering, 119(1): 6990.
Chang-Yu, Ou, Tzong-Shiann Wu & Hsii-Sheng Hsieh,
1996. Analysis of deep excavation with column tye of
ground improvement in soft clay. Journal of Geotechnical
Itasca Consulting Group, Inc. 2002.6. FLAC3D (Fast
Lagrangian Analysis of Continua in 3D Dimensions) User
Manuals, Version 2.1. Minneapolis, Minnesota.
Itasca Consulting Group, Inc.2005.5. FLAC(Fast Lagrangian
Analysis of Continua )User Manuals, Version 5.0. Minneapolis, Minnesota.
Jacob Bear, 1983. Dynamics of Fluids in Porous Media.
Beijing: China Architecture & Building Press.
Lin Li, 2007. Studies on the behavior of deep excavation
and surroundings due to dewatering effect, Ph.D, Thesis,
University of Tongji, Shanghai, China.
Lin Li & MinYang, 2007. The analysis of deformation characteristics of the deep excavation in soft clay. China Civil
Engineering Journal, 40(4): 6672.
Sunil S. Kishnani & Ronaldo I.Borja, 1993. Seepage and soilstructure interaction effects in braced excavtion. Journal
of Geotechnical Engineering, 119(5): 912927.
Youssef M.A. Hashash & Andrew J Whittle, 1996. Ground
movement prediction for deep excavations in soft clay.
Journal of Geotechnical Engineering 122(6): 474486.
Yuqi Li, 2005. Studies on the behavior of foundation pit with
excavation considering seepage, Ph.D, Thesis ,University
of Zhejiang, Hangzhou, China.
152
Analysis of the factors influencing foundation pit deformations

Y.Q. Li
Department of Civil Engineering, Shanghai University, Shanghai, P.R. China
K.H. Xie
Institute of Geotechnical Engineering, Zhejiang University, Hangzhou, P.R. China
J. Zhou & X.L. Kong

Key Laboratory of Geotechnical and Underground Engineering of Ministry of Education,
Tongji University, Shanghai, P.R. China
Department of Geotechnical Engineering, Tongji University, Shanghai, P.R. China
ABSTRACT: Due to the complexity of excavation and groundwater seepage, the behavior of foundation
pits is not yet well understood. In this paper, based on three-dimensional (3D) Biots consolidation theory and
nonlinear Duncan-Changs model, finite element equations considering the coupling of groundwater seepage and
soil skeleton deformation during excavation are deduced and a corresponding three-dimensional finite element
program is developed. Using the program, the influence of soil permeability, rigidity and tiers of supports,
rigidity of retaining wall and construction period of excavation on ground surface settlement, wall horizontal
displacement and pit base heave are analyzed in detail. Some useful conclusions are drawn by analyzing the
influence of these factors on the excavation deformations, which are very significant for guiding design and
construction of excavations.
INTRODUCTION
In urban areas, more and more underground space is

utilized with the fast development of city construction, and thus a lot of excavation engineering appears.
However, the pit deformations induced by excavation greatly influence the safety of not only the pit
itself but also the buildings and municipal facilities
around it. Therefore, study of the behavior of foundation pits has received much attention. Whittle et al.
(1993) described the application of a finite element
analysis for modelling the top-down construction of
a seven-storey, underground parking garage at Post
Office Square in Boston. The results demonstrated that
reliable and consistent predictions of soil deformations
and groundwater flow can be achieved by advanced
methods of analysis without recourse to parametric
iteration, but emphasized the need for adequate characterization of engineering properties for the entire soil
profile. Vaziri (1996) described a simple, efficient and
practical numerical model for analysis of cantilevered
and strutted flexible retaining walls. The model had
incorporated a variety of features that affected the
performance of the retaining walls in the field such
as installation and removal of struts, application of
surcharge, changes in groundwater table, changes in

soil properties and simulation of staged excavations.
The model can be used effectively to perform a broad
suite of parametric studies in the design stage and
also as a reliable tool for predicting performance.
Ou et al. (1996) further proposed a nonlinear, 3D
finite element technique for deep excavation analysis. The technique as well as the analytical procedures
for modeling the excavation processes were coded into
a computer program, and the accuracy of the program
was assessed. The case of an irregularly-shaped excavation with field measurements of wall deflection was
studied and the results showed close agreement with
field measurements. Zdravkovic et al. (2005) studied the effect of excavation on the surrounding areas
and provided a detailed assessment of wall and ground
movements.
There have been a few studies on the influencing
factors of foundation pit deformations. In this paper,
3D consolidation finite element equations are derived,
and the corresponding finite element program is developed. Some useful conclusions are drawn by analyzing
the influence of factors such as soil permeability,
rigidity and tiers of supports, rigidity of retaining
wall and construction period of excavation on the pit
153
deformations, which are beneficial to optimisation of

excavation design.
FINITE ELEMENT EQUATIONS
Based on Biots 3D consolidation finite element equations (Xie & Zhou 2002), and considering groundwater
seepage induced by the water head difference between
the inside and outside of a pit, the finite element
equations of excavation are as follows:
Figure 1. Mesh of finite elements.
Table 1.
of soil.
where is an integral constant; t is the time

increment; [Keij ] and [Kcij ] are respectively the submatrices of the stiffness matrix and the coupling
matrix; Ksij is an element of seepage matrix; ui ,
vi and wi are the displacement increments of element node i; Pi(n+1) is the soil water potential of
element node i at t = tn+1 ; Rxi = Rxi + [Kcij ]Pi(n) ,
Ryi = Ryi + [Kcij ]Pi(n) , Rzi = Rzi + [Kcij ]Pi(n) ,
and Rpi = Rpi tKsij Pi(n) , Rxi , Ryi and Rzi
are the equivalent load increments of element node
i, and Rpi is the equivalent water runoff increment
of element node i, Pi(n) is the soil water potential of
element node i at t = tn .
The soil water potential of a saturated soil can
be expressed using the following equation when the
solute potential of the soil is neglected:
where the spatial coordinate z is upwards positive; P is

soil water potential of saturated soil; p is the sum of the
pressure potential and the load potential, i.e. the total
pore water pressure; and w z is the gravity potential.
3 ANALYSIS OF THE INFLUENCING

FACTORS OF PIT DEFORMATIONS
In order to analyze the parametric influence on the pit
deformations, a 3D consolidation finite element program is developed on the basis of the finite element
equations derived. Using a numerical example given
below, the main factors influencing the pit deformations such as soil permeability, rigidity and tiers of
supports, rigidity of retaining wall and construction
period of excavation are analyzed respectively.
Duncan-Chang model parameters
Parameters
Values
K
n
Rf
c

F
G
D
Kur
150
0.7
0.85
15 kPa
35
0.15
0.35
3.5
300
3.1 Reference case numerical example

The excavated length, width and depth of the foundation pit in a certain homogenous and isotropic stratum
of soft soil are 60 m, 50 m and 8 m respectively. The
soils vertical and horizontal permeability coefficients
are both 2.0 106 cm/s and the effective unit weight
of the soil is 9.0 kN/m3 . The retaining wall is 0.6 m
thick and embedded 16 m deep in soft soil. Reinforced
concrete supports are installed at different excavation
stages and the horizontal spacing between supports
along the pits long side (i.e. y-direction) and short
side (i.e. x-direction) is 6 m and 5 m respectively in
every tier.
In order to minimize the boundary effects and
improve the computational efficiency, the calculation
domains in x-, y- and z-direction are 100 m, 100 m and
40 m respectively in consideration of the symmetry
about the pit centerline. The finite element mesh of
the soil mass and retaining wall are shown in Figure 1.
All soil units are discretized using eight-node
hexahedral isoparametric elements, modelled using
the nonlinear Duncan-Chang model with parameters
listed in Table 1, where c and are the effective
cohesion and the effective friction angle of the soil
respectively, Rf is the failure ratio, and K, n, F,
G, D and Kur are some parameters determined by
tests. The retaining wall adopts Wilson non-harmony
154
3.2
Depth/m
k=2e-7cm/s
k=5e-7cm/s
-8
k=2e-6cm/s
k=5e-6cm/s
-12
k=2e-5cm/s
-16
Figure 2. Influence of soil permeability on wall

displacement.
Distance from the wall/m
0
20
40
60
80
-1
-2
-3
k=2e-7cm/s
k=5e-7cm/s
k=2e-6cm/s
k=5e-6cm/s
k=2e-5cm/s
-4
-5
-6
Figure 3. Influence of soil permeability on ground

settlement.
12
Influencing factors
3.2.1 Soil permeability

In this section, the influence of soil permeability on
the pit deformations at the y = 0 section after the
third excavation stage is studied. The soil permeability for the reference case is 2.0 106 cm/s. Four
more analyses are carried out for soil permeability
of 2.0 105 cm/s, 5.0 106 cm/s, 5.0 107 cm/s
and 2.0 107 cm/s. Figures 24 show the influence
of soil permeability represented by permeability coefficient k on the wall horizontal displacement, ground
settlement and pit base heave. With the soil permeability increasing, the vertical effective stresses outside
the foundation pit also increase, but those beneath the
pit base decrease, so ground settlement and pit base
heave increase, which are shown in Figures 34. For
the wall horizontal displacement, with the soil permeability increasing, the horizontal effective stresses
inside and outside the pit both increase, and the wall
horizontal displacement decreases as a result of the
greater influence of lateral pressures acting on the wall
inside the pit, which can be seen in Figure 2.
-4
Ground settlement/cm
1. Stage 1: 2.0 m excavation depth without supports

for four days, and four days excavation intermission for installing supports at the next excavation
stage. The z value is 1.5 m for the first tier of supports and 2.0 m for the corresponding excavation
level below the supports.
2. Stage 2: 3.0 m excavation depth (excavation to
5.0 m deep) with a tier of supports in six days,
and six days excavation intermission for installing
the next tier of supports. The z value is 4.5 m
for the second tier of supports and 5.0 m for the
corresponding excavation level below the supports.
3. Stage 3: 3.0 m excavation depth (full excavation
to 8.0 m deep) with two tiers of supports in eight
days, and twenty days excavation intermission for
casting the pit base concrete.
Wall horizontal displacement/cm
Pit base heave/cm
elements, modelled as a linear elastic model, whose

modulus of elasticity and Poissons ratio are 25 GPa
and 0.167 respectively. A row of 0.1 m thick interfaces,
connecting the soil mass and the retaining wall is at the
two sides of retaining wall, adopting 3D thin interface
elements derived from Yins rigid plastic model (Yin
et al. 1995) with the outer friction angle = 1.0 and
cohesion = 0.5 kPa, and its other model parameters are
the same as those of the soil mass elements. The supports are modelled using a linear elastic model and
spatial bar elements, with 0.6 m 0.6 m cross section,
whose elasticity modulus is 23 GPa.
The excavation involves three stages. The detailed
description of the staged excavation of the pit is as
follows:
9
6
k=2e-7cm/s
k=2e-6cm/s
k=2e-5cm/s
3
0
10
k=5e-7cm/s
k=5e-6cm/s
15
20
25
Distance from the pit center/m
Figure 4. Influence of soil permeability on pit base heave.
3.2.2 Rigidity of supports

The influence of support rigidity on wall horizontal
displacement, ground settlement and pit base heave
at the y = 0 section after the third excavation stage
are shown in Figures 57. When the support rigidity becomes larger, the retaining wall movement is
more restricted, so the wall horizontal displacement
is smaller. However, the influence of support rigidity
on ground settlement and pit base heave is relatively
insignificant.
3.2.3 Tiers of supports
The influence of support tiers on the pit deformations at the y = 0 section after the third excavation
155
wall horizontal displacement/cm
-8
-12
10
Depth/m
Depth/m
-4
5GPa
10GPa
23GPa
50GPa
100GPa
Figure 5. Influence
displacement.
of
support
rigidity
on
30
40
two tiers
-8
one tier at
upper depth
one tier at
lower depth
no support
-12
20
40
60
5GPa
23GPa
100GPa
-2
80
10GPa
50GPa
-4
20
40
60
80
-3
-6
two tiers
one tier at upper depth
one tier at lower depth
no support
-9
-12
-15
Figure 9. Influence of support tiers on ground settlement.
-6
20
Pit base heave/cm
Figure 6. Influence of support rigidity on ground

settlement.
12
Pit base heave/cm
20
-4
wall
9
5GPa
23GPa
100GPa
6
3
0
10
Figure 8. Influence of support tiers on wall displacement.
-16
-16
10
10GPa
50GPa
two tiers
one tier at upper depth
one tier at lower depth
no support
15
10
10
15
20
25

15
20
Figure 10. Influence of support tiers on pit base heave.
25
Figure 7. Influence of support rigidity on pit base heave.
stage is studied. The reference case has two tiers of

support. Three more analyses are carried out: no support, one tier at 2.0 m excavation depth and one tier at
5.0 m excavation depth. Figures 810 show the influence of support tiers on wall horizontal displacement,
ground settlement and pit base heave respectively. The
deformations of the foundation pit during excavation
with no support are the largest, and they evidently are
smaller with adding support tiers. Tiers of support also
influence the pit deformations, which for two-tier supports are less than those with one-tier. In addition,
the position of the supports also greatly influences

the pit deformations. The deformations with supports
installed at a higher level are less than those with
supports installed at a lower level under the same conditions, so the former approach is more effective for
controlling the pit deformations.
3.2.4 Rigidity of retaining wall
Figures 1113 show the influence of rigidity of the
retaining wall on wall horizontal displacement, ground
settlement and pit base heave at the y = 0 section after
the third excavation stage. The wall horizontal displacement will obviously decrease with an increase
in the rigidity of the retaining wall. However, the
156

3
Depth/m
-4
12
10GPa
-8
25GPa
50GPa
-12
10
t=48d
t=60d
t=72d
-16
-16
Figure 14. Influence of construction period of excavation

on wall displacement.
Figure 11. Influence of rigidity of retaining wall on wall

displacement.
Distance from thewall/m
0
20
40
60
80
0
-1
-2
5GPa
10GPa
25GPa
50GPa
Ground
settlement/cm
t=36d
-8
-12
100GPa
4
t=24d
-4
5GPa
Depth/m
100GPa
-3
-4
20
40
60
-1
t=24d
t=36d
-2
t=60d
t=72d
80
t=48d
-3
-4
-5
-5
-6

on ground settlement.
Pit base heave/cm
Figure 12. Influence of rigidity of retaining wall on ground

settlement.
Pit base heave/cm
12
10
8
6
5GPa
50GPa
2
0
10GPa
25GPa
15
8
t=24d
t=48d
t=72d
6
4
10
t=36d
t=60d
15
20
25
100GPa
10
10
20
25

on pit base heave.
Distance fromthe pit center/m
Figure 13. Influence of rigidity of retaining wall on pit base

heave.
influence of rigidity of retaining wall on ground settlement and pit base heave is not significant. Therefore,
increasing the rigidity of the retaining wall can effectively reduce the wall horizontal displacement and is
beneficial to the safety of excavations.
3.2.5 Construction period
The construction period includes the excavation period
and intermissions at all excavation stages, which is 48 d
in the reference case. Four more analyses were carried
out for construction period of 24 d, 36 d, 60 d and 72 d.
The influence of construction period on wall horizontal displacement, ground settlement and pit base heave
at the y = 0 section after the third excavation stage is
shown in Figures 1416. On the one hand, with the
construction period increasing, the excess pore water
pressures have a longer time to dissipate, and the soil
strata can achieve a higher degree of consolidation,
gaining higher strength and stiffness, thus the wall
horizontal displacement decreases to a certain extent.
On the other hand, the pit base heave increases with
an increase in construction period. The influence of
construction period on ground settlement is relatively
insignificant.
157
REFERENCES
CONCLUSION
Based on Biots consolidation theory, finite element

equations were deduced and a computer program was
developed. The influence of the key parameters such as
soil permeability, rigidity and tiers of supports, rigidity
of the retaining wall, and the construction period on
pit deformations is studied using the finite element
program. The study and the results reported in this
paper are helpful to guide excavation engineering.
ACKNOWLEDGEMENTS
This research project was supported by the China Postdoctoral Science Foundation (No. 20060400672) and
Innovation Fund of Shanghai University, China.
Ou, C.Y., Chiou, D.C. & Wu, T.S. 1996. Three-dimensional

finite element analysis of deep excavation. Journal of
Geotechnical Engineering, ASCE 122(5):337345.
Vaziri, H.H. 1996. A simple numerical model for analysis of
propped embedded retaining walls. International Journal
of Solids Structures 33(16):23572376.
Whittle,A.J., Hashash,Y.M.A. & Whitman, R.V. 1993.Analysis of deep excavation in Boston. Journal of Geotechnical
Engineering, ASCE 119(1):6990.
Xie, K.H. & Zhou, J. 2002. Theory and Application of Finite
Element Analysis in Geotechnical Engineering. Science
Press, Beijing.
Yin, Z.Z., Zhu, H. & Xu, G.H. 1995. A study of deformation
in the interface between soil and concrete. Computers and
Geotechnics 17:7592.
Zdravkovic, L., Potts, D.M. & St John, H.D. 2005. Modelling of a 3D excavation in finite element analysis.
Geotechnique 55(7):497513.
158
Construction monitoring and numerical simulation of an excavation

with SMW retaining structure
Z.H. Li & H.W. Huang
ABSTRACT: The soil mixing wall (SMW) retaining structure is applied in two long strip excavations in
Shanghai. Firstly, the bearing and deforming mechanism of SMW is analyzed in brief. The structural analysis
method of SMW is discussed. Secondly, based on the in-situ excavating construction procedures, the construction
steps of excavating and supporting are simulated in the numerical calculation with the method of Fast Lagrangian
Analysis of Continua 3D. There are two cases simulated in numerical calculation, case 1 is the normal case in
which the supports are installed timely, and case 2 is a case in which the supports are not installed timely because
of some reasons. Then, the deformation of the retaining structure, the horizontal displacement at the top of SMW
and the axial forces of steel pipe supports are analyzed and compared with the actual observation data in two
cases. A good agreement can be found between the calculation results and observation data. It can be seen that
in case 1 the excavation is stable and safe; the axial forces are lower than the alarm values and the displacement
due to excavating is in the permissible range. In case 2, however, the excavation is in danger of instability and
some measures should be taken to protect the SMW retaining structure from failure.
2 THE PROJECT GENERAL SITUATION

AND MONITORING SCHEME
INTRODUCTION
The composite structure with H-shaped steel and

deep cemented-soil piles is called SMW method. This
method can be applied in cohesive soil, sandy soil and
sandy gravel layers. It has been widely accepted in
China, which is mainly applied to deep excavations in
soft soils of eastern and southern China.
This paper studies the bearing and deforming mechanism of SMW. Based on the in-situ excavation construction procedures of the engineering example, the
construction steps of excavating and supporting are
simulated in the numerical simulation method with
FLAC3D. Because the excavations are too long, the
method of excavating is similar to tunnel excavating,
which is from one side to another. And there are two
cases simulated, one is the normal case in which the
supports are installed timely. However, in middle of
March 2007 during excavating, because of the bad
weather and some other reasons, the supports were not
installed timely so that the excavation was in danger
of collapse. Therefore, another case is a case in which
the supports are not installed timely.
Through comparing the actual observation data
in-situ and the calculation results, some useful conclusions are achieved.
2.1 The project general situation

The underground channels project contains two long
strip excavations, the eastern and western excavations,
which are similar to each other. The length is both
820.5 m, and the width is 16.5 m. The excavation depth
is between 6.645 m 8.039 m. It is clear around the
construction site.
The foundation soil layers belong to Quaternary Pleistocene-Holocene deposit, including cohesive soils, silt and sandy soils which distribute in
planes. The physico-mechanical parameters of soils
in site are shown on Table 1.
The 650 SMW method is applied and steel pipe
supports are installed. The SMW retaining structure is 17 m long in depth, which is inserted with
500 200 10 16 H-shaped steels. In order to facilitate the earthwork excavating, two steel pipe support
tiers including the upper and the lower supports are
installed considering the characteristic of the excavations. The upper steel supports locate 1.0 m below the
ground surface. The lower steel supports locate 2.5 m
above the excavation bottom. The sizes of the steel
159
Table 1.
The soil
layers
includes the upper EZC1 and the lower EZC1 supports.

Therefore, there are 12 monitoring points for every
excavation, EZC denotes the axial force monitoring
points in the eastern excavation, and WZC denotes the
western one. The distance between each pair monitoring points is about 120 m. The steel strain gauges
are welded around the steel pipe supports as shown in
Figure 1(b).
Physico-mechanical parameters of soils in site.

Depth
C
(m)
(kN/m3 ) (KPa) ( )
Silty clay
3
and clayey
silt
Sandy silt
2
Silty clay
2
Sandy silt
2
and silty
clay
Silty clay
10
Clay
11
E
(MPa)
17.8
10.0
23.0
4.38
0.3
18.2
17.3
17.6
4.0
10.0
3.0
29.0
21.5
25.5
4.78
3.84
5.08
0.33
0.35
0.3
16.2
17.3
11.0
13.0
11.0 13.62
12.0 17.0
0.3
0.3
Figure 1. General layout of monitoring points: (a) The south

section of the excavations; (b) The strain gauges welded
around the support.
supports are all 609 12. The distance between the

adjacent supports is 5 m.
2.2 The monitoring scheme
Around the two excavations, the monitoring points for
the horizontal displacement and settlement at the top
of the retaining structure are located about every 10 m.
And they are numbered using E and W, in which E
denotes the eastern excavation, W denotes the west
excavation. There are all 324 monitoring points for
the top horizontal displacement and settlement of the
SMW in two excavations, as shown in Figure 1(a).
Because the excavations are so long that the south
section is given only.
14 inclinometer tubes for lateral deformation of
SMW are set in every excavation. They are located
in the same distance about 100m symmetrically.
The location of every monitoring point is shown in
Figure 1(a), in which ECX denotes the inclinometer
tube in the eastern excavation and WCX denotes the
inclinometer tube in the western one.
In every excavation there are 6 pairs of monitoring points for axial forces of supports, and every
pair has two points including the upper and the lower
supports. As shown in Figure 1(a), for example, EZC1
BEARING AND DEFORMING

MECHANISM OF SMW
The cemented soil material that is produced generally has a higher strength, lower permeability, and
lower compressibility than the native soil. Therefore,
the SMW method can make it possible to form waterpreventing and earth-retaining walls quickly by mixing
earth collected at a construction site with cement
slurry. The rigidity of the earth retaining walls was further enhanced by forming a compound earth-retaining
wall with H-shaped steel materials welded with studs
that act as stress material arranged within the improved
soil walls. And under the suitable conditions, the
H-shaped steels can be recycled.
Stress-strain characteristics of SMW are extremely
complex during the course of the pit excavation. The
curves of H-shaped steel strain are under the linear
elastic scope, but cemented-soil is nonlinear response,
and the rigidity changes of composite structure mainly
by the cemented-soil. It is commonly considered that
the H-shaped steels bear all the lateral water and earth
pressure and the cement deep mixing piles are used to
prevent water. However, it is testified through experiments that cement soil can enhance the H-shaped steels
to reduce the deformation. In addition, the cement
soil can also have confinement effect to prevent the
H-shaped steels instability. The composite flexural
stiffness is 20% greater than only H-shaped steels. The
stiffness enhancing coefficient can denote the degree
of stiffness enhancing as follows:
where Ecs and Es are the elastic modulus of cement

deep mixing pile with H-shaped steel and the elastic
modulus of H-shaped steel, respectively; Ics and Is are
the inertia moment of cement deep mixing pile with
H-shaped steel and the inertia moment of H-shaped
steel.
In this numerical calculation, the cement deep
mixing pile with H-shaped steel is equivalent to
diaphragm wall and the influence of stiffness enhancing coefficient is considered. According to the
principle that the stiffness is equal to each other, the
equation is given by
160
Figure 2. The stiffness equivalence between SMW and

diaphragm wall.
second layer is from 2.0 m to 6.5 m, the third is

from 6.5 m to 8.0 m. There are upper and lower
two supports installed, the upper supports are located
at 1.0 m and the lower supports are at 6.5 m. And
the distance between two adjacent supports in y direction is 5 m. Therefore, the length of soils excavated in
every layer is 5 m in y direction in every excavating
step. Because the excavation is too long, the excavating
method is similar to tunnel excavating method which
is from one side to another. The construction procedure of excavating and supporting is divided into lots
of steps, as follows:
The construction of SMW.
The first layer is excavated 5 m in y direction and
the first upper steel support is installed.
The first and second layers are excavated 5 m in y
direction and the second upper and the first lower
supports are installed.
All the three layers are excavated 5 m in y direction,
and the third upper and the second lower supports
are installed.
Do this until the earthworks excavation is
completed.
Figure 3. Elements of model in case 1.
Based on equation (1), thus
Here, is considered as 1.2. The equivalent thickness of diaphragm wall in this numerical calculation is h = 0.65 m and the Young modulus is
E = 12.6 GPa.The equivalent figure from SMW with
H-shaped steels to diaphragm wall in this project is
shown as Figure 2 shown.
The interfaces are installed to simulate the interface characteristics between the retaining structure
and the soils. In FLAC3D, Interfaces have the properties of friction, cohesion, dilation, normal and shear
stiffness, and tensile and shear bond strength, which
are characterized by Coulomb sliding and/or tensile
and shear bonding. In this computation, the equivalent
diaphragm wall is considered as elasticity.
4
CALCULATION CASE
4.1 Case 1
Because there are two long strip excavations and they
are similar, one part of the eastern excavation is chosen
to be simulated. The model size is 60 m in extent, 60 m
in breadth and 30 m in height. The model is shown in
Figure 3.
The earthworks soils for excavating are divided into
3 layers. The first layer is from 0.0 m to 2.0 m, the
161
The whole procedure of excavating steps and

installing supports is simulated by 3D numerical
method. There are 11 excavating and supporting steps
except the construction of SMW.
In this numerical simulation calculation, the
mechanical soil behavior is modeled with MohrColumn model and the supporting structures are considered as elastic model. The interfaces are installed
between SMW and soils. The top of the model, at
z = 30 m, is a free surface. The base of the model, at
z = 0 m, is fixed in the z-direction, and roller boundaries are imposed on the sides of the model, at x = 0 m,
x = 60 m, and y = 0 m, y = 60 m.
4.2
Case 2
In middle of March 2007, because of the bad weather

and other reasons, the steel pipe supports were not
installed in time. There were about 30 m in length without supports from the excavating face to the lattermost
supports for a long time. Meanwhile, according to the
in-situ measurements, there was a sharp increment in
horizontal displacement of the soil mixing wall. This
case is simulated to analyze the influence.
In this case, the material properties and boundary
conditions are same to case 1. The excavating and
supporting procedures are same, too.
5
5.1
RESULTS
Case 1
In order to analyze the calculation results conveniently,

some key points are set in the model, as shown in
Figure 4. The horizontal displacement at point A in case 1.
Figure 5. The horizontal displacement of the SMW in line

B in case 1.
Figure 3. Firstly, the horizontal displacement at the top

of SMW is analyzed. The curve of horizontal displacements at point A is shown in Figure 4(a). It can be seen
that the horizontal displacement increases before step
6 but then decreases in the following steps. The maximum value is 6.76 mm at step 6 and the ultimate value
is 4.63 mm. According to the excavating steps in calculation, when the excavating face exceeds point A about
15m, the value of the horizontal displacement begins
to decline. The actual observation data for point A is
shown in Figure 4(b). It can be seen that the actual values are bigger than the calculation results. The curve
is monotone increasing by steps and tends to be constant after having reached a certain level. Its ultimate
value is 10.4 mm. The calculation result of the horizontal displacement at point A is much less than the
observation value.
The curves of calculation results with excavating
steps and the observation data with date for the horizontal displacement of the retaining structure in line
B are shown in Figure 5. According to the calculation
results, its maximum horizontal displacement of the
SMW occurs at the point of 6.5 m depth, and its value is
22.35 mm. As shown in Figure 5(a), when the excavating face reaches the line B at step 3, the deformation of
SMW increases dramatically. When it has passed away
from line B about 20 m, the deformation increases
slowly. Figure 5(b) shows the observation curve of horizontal displacement of SMW in line B with the date.
A rather good agreement can be found between (a) and
(b). According to the observation data, the maximum
horizontal displacement in line B occurred at point of
5.0 m depth and its value is 26.8 mm, which is greater
than the maximum calculation value by 4.45 mm. With
the excavating face advancing, the deformation increment is becoming smaller. From the actual observation
data and calculation results, it can be seen that the horizontal displacement of the SMW mainly occurred during the period of excavating surface passing this line.
The curves of axial forces of the steel pipe supports C and D with excavating steps are shown in
Figure 6(a). The final axial force of the upper support
C is 657.60 kN, and the lower support D is 1467.7 kN
in calculation results. As shown in Figure 6(b), the
observation data is greater than the calculation results.
The maximum axial force of the upper support C is
995.83 kN, and the lower support D is 1575.7 kN. Both
C and D have an ascending firstly and then declining
process with lapse of time in actual observation. This
is because the foundation mat bore a part of soil pressure with its pouring and strengthening. However, the
procedure of pouring and strengthening of foundation
mat is not simulated in numerical calculation. So there
162
Figure 7. Elements of model in case 2.
Figure 8. The horizontal displacement and the axial forces

of steel pipe supports in case 2.
Figure 6. The axial force of supports C and D in case 1.
is no declining trend of axial forces. But in the early

stage, the trend and shape of calculation results curve
with excavating steps and observation data curve with
date is agreed generally. As shown in Figure 6(b), on 28
April the lower support D was removed, therefore, the
axial force of the upper support C had a significant rise
by 102.5 kN. However, in calculation this procedure is
not simulated.
Through analysis, it can be seen the values of
observation data are greater than the calculation
results universally. The main discrepancy between
calculation and measure can be explained that the
physico-mechanical parameters of soils are not accurate enough. However, in reality, this area was less stiff
than initially planned. According to the numerical simulation and the actual observation data, the excavation
is stable if the steel pipe supports can be installed in
time. The numerical results and actual data of axial
forces are lower than the alarm values. The displacement due to excavating is in the permissible range.
5.2
Case 2
In case 2, the numerical calculation model is shown

in Figure 7. Figure 8 shows the horizontal displacement of the model and the axial forces of steel pipe
supports at last step. It can be seen the maximum horizontal displacement at the top of SMW occurs at point
Figure 9. The horizontal displacement of point E in case 2.
E with 48.96 mm, as shown in Figure 7. The curve

of horizontal displacements at point E when excavating from step 1 to step 11 is shown in Figure 9. It
can be seen the displacement at point E is nearly zero
before step 6 until the excavating face passes point
E. With the excavating face advancing after step 6,
the displacement increases significantly. According to
calculation results, the maximum horizontal displacement of SMW occurs at the point of 3 m depth, which
is in line F, and its value is 50.44 mm. As shown in
Figure 10, the horizontal displacement of SMW in line
F develops slightly until the excavating face passes this
line at step 6. Because the supports are not installed
near this line, the deformation of SMW develops more
rapidly. In actual observation, there are three points of
which horizontal displacements in line F are greater
than the alarm value with 50 mm. The axial forces
163
for this excavation because the supports were not

installed in time. The risk was existent so the corresponding measures should be taken. After being
alarmed, the construction team stopped excavating
and installed the supports speedily. Its turned out
that the measures are very effective according to the
subsequent observed data.
6
Figure 10. The horizontal displacement of the SMW in line
F in case 2.
In this paper, firstly, the bearing and deforming mechanism of SMW is analyzed in brief; secondly, the
construction monitoring scheme is introduced; thirdly,
a 3D numerical simulation of this long stripe excavation is described, including all the components of
the project (the SMW construction, stepped excavation
and supports installation); then the numerical results
are compared with the actual data in-situ observation.
The 3D numerical method can simulate the whole
excavation construction very well. A good agreement can be found between the numerical results
and the actual observation data except for some
small deviations.
The excavation is stable and the displacement due
to excavating is in the permissible range if the steel
pipe supports are stalled timely. However, because
of bad weather and other reasons the steel pipe supports are not installed in time and without supports
for a long time, such as case 2, the excavation is in
danger of collapse.
According to the in-situ observation data, the construction team can take corresponding measures to
protect the excavation away from some undesirable
events and risks.
Figure 11. The axial forces of supports G and H in case 2.
of lastly installed supports G and H are much larger

because of without installation of subsequent supports.
The axial force of support G is 1924.50 kN, and H
is 2095.60 kN. The curves of the axial forces of supports G and H with the excavating steps are shown in
Figure 11. According to calculation results, the axial
forces of G and H are much greater than the adjacent supports by 903.75 kN for upper support and
546.00 kN for lower support. Compared with Figure
6(a), the magnitude is much greater. The axial force
of G exceeds the alarm value with the upper supports
for 1500kN and H exceeds the alarm value of lower
supports for 2000 kN.
There are two inclinometer tubes ECX2 and ECX13
for lateral deformation of SMW around this site.
According to the observation data of 19 March
and 20 March, the velocity of horizontal displacement exceeded the alarm value of 3 mm/d for two
days between the depth of 5 m 8 m at ECX13.
The velocity values were 4.18 mm/d, 3.97 mm/d,
3.64 mm/d and 3.20 mm/d when the depths are 5 m,
6 m, 7 m and 8 m respectively on 19 March. Moreover, the velocity values were 5.18 mm/d, 5.66 mm/d,
4.98 mm/d and 3.44 mm/d in the next day. Meanwhile,
the velocities of horizontal displacement were greater
than 3 mm/d in the depth from 1 m to 12 m at ECX2 on
19 March, which were over 10 mm/d within 3 m from
the top of SMW.
According to the results of numerical simulation
and actual observation data, it was possible to collapse
CONCLUSIONS
REFERENCES
Commend, S. Geiser, F. & Crisinel, J. 2004. Numerical simulation of earthworks and retaining system for a large
excavation. Advances in Engineering Software Vol. 35:
669678.
Li, J.C. Zhang, Z.Y. & Xu, Q 2005. Study on threedimension numerical simulation of deformation of the
deep-foundation pit with excavation. Journal of Nanjing
University of Technology 27(3): 17.
Liu, H.Y. et al 2006. Numerical analysis on excavation
safety of deep foundation engineering. Chinese Journal
of Geotechnical Engineering 28 (Supp): 14411444.
Liu, J.G. & Zeng, Y.W. 2006. Application of FLAC3D to
simulation of foundation excavation and support. Rock
and Soil Mechanics 27(3): 505508.
Liu, J.H. & Hou, X.Y. 1997. The Handbook of Foundation Engineering, Beijing: China Architecture & Building
Press. 569572.
Zhang, P & Liu, R.H. 2000. The Application of SMW Method
in Foundation Pit. Chinese Journal of Rock Mechanics and
Engineering 19(Supp): 11041107.
164
A simplified spatial methodology of earth pressure against

retaining piles of pile-row retaining structure
Y.L. Lin
Geotechnical Research Institute, Hohai University, Nanjing, P.R. China
Key Laboratory for Geotechnical Engineering of Ministry of Water Resource, Hohai University, Nanjing, P.R. China
X.X. Li
ABSTRACT: Pile-row retaining structure is widely adopted in the excavation of deep foundation pit. In the supporting system, retaining piles are the main bearing members. It is extremely important to obtain the magnitude
and distribution of earth pressure against retaining piles. Based on the mode of failure, a new methodology is proposed to evaluate the earth pressure against retaining piles of pile-row retaining structure. In the proposed method,
both spatial effect and intermediate principal stress effect are considered. Finally, the methodology is applied to
practice engineering. It is demonstrated that the strength theory has more influence on earth pressure and the
potential strength of filling materials is sufficiently developed under the guidance of the united strength theory.
INTRODUCTION
The pressure against the back of a retaining structure

caused by backfill and surcharge on the ground surface
is a classical problem of soil mechanics. It is influenced by retaining structure types, movement mode,
stiffness and contact conditions between soil and structure (Fang & Ishibashi, 1986; Harrop-Willrams,1989;
Zhou,1990; Fang et al., 1994; Wang, 2000; Pal &
Salgado, 2003). In addition, the distortion of soil mass
has a certain effect on the earth pressure. Before the
soil achieves breakage, the magnitude of earth pressure
cannot be determined. Even if it reaches limit state, the
earth pressure cannot also be calculated because inner
soil mass cannot synchronously arrive at limit equilibrium state. So the reliable parameter of soil cannot be
acquired. Thus, to apply in practice expediently, it is
usually assumed the soil is on ideal failure state.
In the excavation of deep foundation pit pile-row
retaining structure is widely adopted. In the supporting
system, retaining piles are the main bearing members. It is very important to obtain the magnitude and
distribution of earth pressure against retaining piles.
Because of failure mechanism of soil behind piles,
the influence of interaction between retaining structure
and soil on earth pressure cant be achieved accurately
according to classical earth pressure theory. So earth
pressure should be taken as the spatial problem rather
than plane problem. Mohr-Coulomb strength theory is

usually introduced into the computation analysis and
the influence of intermediate principal stress is omitted. However, plenty of experiments reveal the soil
strength varies with the intermediate principal stress
(Yu, 2004), which is quite different from what has been
depicted in the conventional Mohr-Coulomb theory.
The unified strength theory is a system of yield and
failure criteria of material sunder complex stresses. It
has a clear physics and mechanics background, a unified mathematical model, and a simple and explicit
criterion which includes all independent stress components and simple material parameters (Yu, 2002).
In this study, a new methodology based on the
plastics limit analysis is proposed to evaluate the earth
pressure against retaining piles of pile-row retaining
structure based on the mode of failure. In the proposed
method, both spatial effect and intermediate principal
stress effect are considered. The solution of the equation is obtained, giving a theoretical result for the earth
pressure on retaining piles.
2
UNIFIED STRENGTH THEORY
Based on a twin-shear element and the multiple slip

mechanism, Yu and He (1992) established the unified
strength theory. It has a unified model and simple
165
where c0 = cohesion, 0 = friction angle, b = unified

strength parameter that reflects the influences of the
intermediate principal stress on the yielding of the
material (0 b 1), t and c are uniaxial tensile
strength and compressive strength, respectively, and
is tensile-compressive strength ratio.
Introducing Lode parameter , thus
Substituting Equation (5) into Equation (3) and (4),

letting
The unified cohesion ct and the unified friction

angle t can be defined as
Figure 1. Varieties of the unified strength theory on deviatoric plane (Yu, 2004).
unified mathematical expression that is suitable for

various materials (Yu, 1994). The unified strength
theory covers all the regions from the lower bound
to upper bound, as shown in Figure 1. The unified
strength theory considers the different contributions
of all stress components acting on the stress element
to the yield or failure of materials.
The mathematical modeling is expressed as follows
(Yu et al., 2002):
According to Mohr circularity of stress state at a

point, the unified expression of shear strength can be
obtained
If it is prescribed that press stress is positive,

Equation (1) and Equation (2) can be rewritten as
FAILURE MODE OF SOIL HEHIND PILES
For pile-row retaining structure, the arching effect in

the retaining soil mass occurs (Hu et al., 2000). It
is a stress redistribution process by which stress is
transferred around a region of the soil mass, which
then becomes subject to lower stresses. So the earth
pressure acting on piles is enhanced, while the earth
pressure on soil around piles is depressed. The smaller
is pile space, the stronger is soil arching effect. And it
is more propitious to the stability of foundation pit.
3.1 Simple shear failure mode
Figure 2 shows the failure mode A homogeneous foundation pit of depth H and the net space B is considered.
166
Figure 2. Simple shear failure mode of soil mass between

piles.
Figure 4. Schematic for earth pressure of simple shear

failure.
Figure 3. Rip-shear failure mode of soil mass between piles.
The soil wedge is assumed to be rigid and slide alone a

planar surface. The critical inclination of failure plane
is expressed as .
When B 2b 0, simple shear failure mode occurs,
letting
region soil of block BCDB C D . It is assumed that the

soil is perfectly plastic and their deformation is governed by the associative flow rule. Then, Kinematical
admissibility requires the velocity jump vector, v, be
inclined to the velocity discontinuity at angle t , as
presented in Figure 4(b).
The Cartesian coordinates system established for
the present analysis is shown in Figure 4(a). The
point C is the origin of the Cartesian coordinates system, and plane XY is level plane. Section AC (or A C )
is slippage tangent of fracture surface BCC B , and it is
inclined to the velocity discontinuity at angle t . Slimily, the velocity of fracture surface BCD and B C D is
at an angle t to the slippage tangent DC and D C . In
the plane BCD, according to the directional derivative
of the velocity v, the angle can be obtained as
According to Equation (10), the failure condition of

simple shear failure mode can be obtained as
where Hcr = critical height of failure mode.

3.2
Rip-shear failure mode
Rip-shear failure mode will arise when H > Hcr , as

shown in Figure 3.The soil mass behind piles is divided
into two portions from the critical height Hcr . Above
the height Hcr , the soil is rip failure, and it is shear
failure below the height Hcr .
4
4.1
The rate of work of soil weight can be calculated as

the work rate of block ABC-A B C plus the work rates
for blocks C-BAD and C -B A D . Consequently, this
work rate takes the form
where = the soil unit weight, d = the diameter of

pile, f1 is a function, it can be determined as
CALCULATION OF ACTIVE EARTH

PRESSURE
Earth pressure of simple shear failure mode
Figure 4(a) shows the mechanism of simple shear failure. The single pile endures the earth pressure of the
Due to the homogeneous soil masses being rigid, the

internal energy is only dissipated along the sliding surface. The work dissipation rate can be calculated as the
167
work dissipation rate of block CC DD plus the work

dissipation rates for blocks DBC and D B C . Consequently, this dissipation work rate can be calculated as
where
Since soil-pile interface can be considered as velocity discontinuity rather than stress characteristic, ideal
plastic model is not applicable to the interface material. The relative movement between soil and pile,
which depends on the interface characteristic sand the
property of the adjacent soil, is not always of purely
frictional sliding. If it is assumed that the total horizontal earth pressure induced by soil mass is Pau . The
friction angle of soil-pile interface is . Two conditions
are considered in the following.
1 Smooth pile ( < t )
The external work rate contributed by the resultant
horizontal earth pressure Pau is
The work dissipation rate along pile surface is
Figure 5. Schematic for earth pressure of rip-shear failure.
4.2 Earth pressure of rip-shear failure mode

The rip-shear failure mechanism for the present analysis is shown in Figure 5. The earth pressure acting on
single pile is induced by the region soil mass FCDGG D C F . Similar to the earth pressure of simple shear
failure mode, the rate of work due to the soil mass
weight can be expressed as
Equating the rate of internal energy dissipation to

the rate of the external work, we can obtain
where
Rough pile ( t )
The external work rate contributed by the resultant
horizontal earth pressure Pau is
The work dissipation rate along pile surface is
where S = the center distance of two adjacent piles.

For the rigid material considered, the internal
energy is only dissipated along the sliding surface
and the interface surface of soil-pile. The rate of
energy dissipation along the sliding surface can be
expressed as
Similarly, equating the rate of internal energy dissipation to the rate of the external work, the resultant
horizontal earth pressure can be obtained
168
Figure 6. Relations between the unified strength parameters and the soil strength.
where
(25) with respect to , and equating it to zero and

solving it, we can obtain the critical angle cr . Substituting cr into Equation (25), we have the maximal
upper-bound for the active earth pressure.
The rate dissipation along the interface surface of

soil-pile is similar to the Equation (16) and (19).
For Smooth pile ( < t ), according to the energy
conversation law, we can obtain
Similarly, for rough pile ( t ), the total horizontal

earth pressure Pau can be derived as
Accordingly, the corresponding resultant Pa acting

on the pile is
The magnitude of active earth pressure can be

obtained from Equation (25). Obviously, for a given
example, the resultant Pa is only determined by failure angle . By taking the first derivatives of Equation
RESULTS AND DISCUSSION
To evaluate the validity of the proposed method, a pilerow retaining structure without anchor is analyzed.
Numerical results are presented and compared.
5.1
Effects of the unified strength parameters on

soil parameters
The influences of the unified strength parameters on

soil mass strength are represented in Figure 6 for
0 = 15 . It is observed that the soil mass strength
varies with the variety in the value of the unified
strength b. The soil mass parameters of ct and t are
piece-wise functions and they achieve extremum when
= sin0 .
5.2 Effects of the unified strength parameters on
failure mode
Figure 7 presents the effects of the unified strength
parameters on critical height Hcr for = 17 kN/m3 ,
0 = 15 , H = 10 m. As a whole, the critical height
Hcr decreases with the increase in the value of b.
Figure 7(a) and Figure 7(b) also show the influence
of S and pile diameter d. For c0 = 0, the critical height
increases with the increase in the value of S, while
it decreases with the increase of d. It is also clear
169
Figure 7. Influence of the unified strength parameters on critical height Hcr .
that under the same conditions Hcr decreases with the

increase of c0 or in Figure 7(c) and Figure 7(d).
Table 1. Effects of the unified strength parameters on active

earth pressure.
5.3 Effects of the unified strength parameters on

active earth pressure
Deg
(1)
(2)
(1)
(2)
(1)
(2)
0
15
483.6
364.1
483.6
364.1
414.9
305.4
396.3
378.8
378.8
275.5
350.0
252.0
b=0
Based on the unified strength theory, the values

of active earth pressure is shown in Table 1 for
= 17 kN/m3 , 0 = 15 , H = 10 m, d = 1.0 m, c0 = 0.
The proposed formula can be degenerated into the
expressions induced by Mohr-Coulomb strength theory. From the table, it is found that the strength
170
b = 0.5
b=1
b = 0 is simplified to Mohr-Coulomb theory; b = 1 is twin

shear theory. Situation (1) represents = 0, and situation
(2) is = sin0.
theory has more prodigious influences on the active

pressure. The resultant earth pressure Pa decreases
with the increase in unified strength parameter b. For
= 0, when b changes from 0 to 1, the resultant
earth pressure Pa decreases by about 21.7%. Similarly, the earth pressure Pa is also influenced by the
Lode parameter .
6
CONCLUSIONS
The estimation of active earth pressures acting on

retaining piles is very important in geotechnical
design. However, unlike the assumption used in the
analysis of Coulomb, which generally calculates earth
pressure according to plane strain, the earth pressure
behind the piles should be taken as the spatial problem. This is due to arching effects in the retaining soil,
which result from the frictional resistance between the
piles and the soil.
In this paper, a new methodology is proposed to
evaluate the earth pressure against retaining piles of
pile-row retaining structure. The advantage of the proposed method lies in the fact that both spatial effect and
intermediate principal stress effect are considered. It is
indicated that the strength theory has more influence
on earth pressure and the potential strength of filling
materials is sufficiently developed under the guidance
of the united strength theory. But the methodology
requires experiment or field verification.
Fang,Y.S., Chen,T.J. & Wu, B.F. 1994. Passive earth pressures

with various walls movements. Journal of Geotechnical
Engineering, ASCE, 128(8):651659.
Harrop-Willrams, K.O. 1989. Geostatic wall pressures. Journal of Geotechnical Engineering, ASCE, 115(9):1321
1325.
Hu, M.Y., Xia, Y.C. & Gao, Q.Q. 2000. Calculation principle of earth pressure against retaining piles of pile-row
retaining structure. Chinese Journal of Rock Mechanics
and Engineering, 19(3):376379.
Palk, K.H. & Salgado, R. 2003. Estimation of active earth
pressure against rigid retaining walls considering arching
effects. Geotechnique, 53(7):643653.
Wang,Y.Z. 2000. Distribution of earth pressure on a retaining
wall. Geotechnique, 50(1):8388.
Yu, M.H. 2004. Unified strength theory and its applications.
Berlin: Springer.
Yu, M.H. 2002. Advance in strength theory of materials under
complex state in the 20th Century. Applied Mechanics
Reviews, 53(3):159218.
Yu, M.H. 1994. Unified strength theory for geomaterials
and its applications. Chinese Journal of Geotechnical
Engineering, 14(2):110.
Yu, M.H., He, L.N. & Liu, C.Y. 1992. Generalized twin
shear stress yield criterion and its generalization. Chinese
Science Bulletin, 37(24):20852089.
Yu, M.H., Zan, Y.W. & Zhao, J. 2002. A unified strength
criterion for rock material. International Journal of Rock
Mechanics and Mining Sciences, 39:975989.
Zhou, Y.Y. & Ren, M.L. 1990. Experimental study of the
active earth pressure on rigid retaining wall. Chinese
Journal of Geotechnical Engineering, 12(2):1926.
REFERENCES
Fang, Y. & Ishibashi, I. 1986. Static earth pressures with
various wall movements. Journal of Geotechnical Engineering, ASCE, 112(3):317333.
171
Consideration of design method for braced excavation based on

monitoring results
H. Ota, H. Ito & T. Yanagawa
Osaka Municipal Transportation Bureau, Osaka, Japan
A. Hashimoto
Kotsu Service Co., Ltd., Osaka, Japan
T. Hashimoto & T. Konda

ABSTRACT: A comparison between observed data and design value of earth retaining wall deflection due to
braced excavation was carried out in soft and sensitive clay ground of some construction sites of Osaka Subway
Line No.8. The beam-spring model was employed in the braced design method, and it was taken into account the
characteristics of the Osaka soft ground, and there was good agreement between the observed data and design
value in past results. According to this comparison, the observed wall deflection was larger than the designed
one in some construction sites consisted of the soft and sensitive clay layer with 10 to 20 m thickness. In this
paper, the measuring process of the horizontal coefficient of subgrade reaction kh in the excavation side of soft
clay layer is discussed. As the value of kh became small depended on the wall deflection, the new knowledge has
been employed on the design method. It is found that the calculation with the revised design method agree well
with the monitoring data.
INTRODUCTION
for design has been described based on the ground

properties.
In densely populated city, it is necessary to use the

underground space highly and effectively for the development of city. It is believed that the demand for
much deeper underground excavation will increase
gradually. Therefore an applicable design method is
demanded for deep, safe and economical excavation.
Osaka MunicipalTransportation Bureau (OMTB) suggested an original design method for braced excavation based on the characteristics of the Osaka
ground and subway constructions. At each construction site (elevens stations and railway depot) where
open cut method was adopted in Osaka Subway Line
No.8, braced excavation design based on this original
design guideline was carried out, and the observational
method was also utilized effectively.
In this paper, some comparisons between observed
data and design value of earth retaining wall deflection due to braced excavation have been carried out on
soft and sensitive clay ground of two construction sites
of Osaka Subway Line No.8. The evaluation method
CHARACTERISTICS OF THE MODIFIED

BEAM SPRING MODEL
East and West sides of Osaka ground are consistent

with the Holocene layers (soft clay and loose sand), but
Pleistocene layers exist around the ground surface of
Uemachi plateau. The water levels are high in unconfined and confined aquifers, also the permeability of
these aquifers are large.
The beam-spring model for the braced excavation,
which is indicated in Standard Specifications for Tunnel [Open Cut Tunnel] published by the Japan Society
of Civil Engineers (JSCE, 2006), is frequently implemented as a widely usable method in Japan. However,
since the result of the prediction of wall deflection and
strut force are not always consistent with the observation data, OMTB proposed the modified beam spring
design method (OMTB, 1993) (Kishio et al., 1997)
173
which can consider the characteristics of Osaka

ground and the conditions of subway construction.
The outline of the OMTB model is shown in Figure 1.
2.1 Active lateral pressure above the excavation
bottom
Because there are some possibilities of gap between
the braced wall and back ground, the active earth
pressure is estimated by Rankine-Resals equation with
the water pressure. In sandy layer, the water pressure
is assumed as hydrostatic pressure. In clayey layer, the
water head of the upper sandy layer is extended in this
clayey layer.
2.2 Active lateral pressure below the excavation
bottom
In the case of sufficient penetration depth of braced
wall, the wall deflection near the tip is small and
the lateral pressure is kept as the initial condition.
So, if the active lateral pressure is defined basically
only by the limit equilibrium theory, there are some
cases which the wall deflection is overestimated by
giving much lateral pressure. Therefore, the active lateral pressure which is deeper than the bottom layer of
excavation is gradually decreasing in the area of triangle formed from the lateral pressure at the bottom of
the excavation to the tip of wall.
2.3
Resisting lateral pressure of the excavated

ground
Resisting lateral pressure of the excavated ground

is the multiplication of the coefficient of horizontal
subgrade reaction and the wall deflection. However,

this value should not exceed the coefficient of the
passive lateral pressure which is the subtraction from
limit passive lateral pressure defined by Coulombs
equation to the lateral pressure at rest.
2.4 Water pressure in clayey ground

Since it is difficult to distinguish the water pressure
from the lateral pressure in clayey ground, lateral pressure is often identified as the integration of soil and
water. On the other hand, it is considered that soil is
separated from water in modified beam-spring model.
Because the pore water pressure acts on the braced
wall hydrostatically, the water path is possibly formed
between the wall and the back ground due to braced
excavation.
For these reasons, the effective stress method is
adopted in both sandy layer and clayey layer. The
groundwater table in clayey ground is taken on the
higher gravitational pressure distribution by comparison between the upper water-bearing layer and down
side water-bearing layer.
2.5 Supported effect of covering plates

Because the effect of depressing the wall deflection is
recognized when the covering plates are constructed
in the same direction as struts, the supported effect of
covering plates is considered by reducing 10% of the
spring-beam coefficient.
2.6 Horizontal coefficient of subgrade reaction of

excavated side
The coefficient of subgrade reaction kh used in the
JSCE model is taken into consideration the geometrical effect related to the difference of loaded
width based on some plate loading test results performed near the ground surface. However, the lateral
pressure acts on the horizontal direction against the
earth retaining wall, because the wall is installed
to the vertical direction in subway construction site.
Therefore, it is not always appropriate to apply the
coefficient of subgrade reaction used in the JSCE
model to braced excavation design directly. So in the
OMTB model, the coefficient of subgrade reaction
is expressed as equation (1) and (2) (Yanagida et al.,
1981) empirically.
Figure 1. The concept of the modified beam spring model

(presented by OMTB, 1993).
174
COMPARISONS BETWEEN OBSERVATION

AND DESIGN OF BRACED WALL
DEFLECTION
3.1 A-site
The comparison between observed data and design

value of earth retaining wall deflection due to braced
excavation was carried out in soft and sensitive clay
ground of No.8 Line construction sites.
In general, the design value can estimate the
observed data appropriately in most construction sites.
But in some sensitive and soft alluvial clay layer accumulate from 10 m to 20 m thick, observed data are
larger than the design value due to braced excavation. The causes for these phenomena are described
as follows.
Figure 2. Cross section (A site).
Table 1.
The layer of this A-site ground consists of the alluvial

layer, upper lower Pleistocene from the ground surface.
The uniformity coefficient of this fine sand Aus
is small, and is called the first water-bearing layer.
The alluvial clayey layer Auc is sensitive (N-value = 0
to 3, liquid limit IL = 0.4 to 1.0, cohesion c = 20 to
100 kN/m2 ), and is typical soft ground in this construction site. On the other hand, under the alluvial layer,
the upper Pleistocene sandy and gravel layer Tsg (the
second water-bearing layer), the lower Pleistocene clay
layer Oc3 (Osaka Group, Marine Clay Ma3, c = about
400 kN/m2 ) and the lower Pleistocene sandy layer
Os3 (Osaka Group, the third water-bearing layer, Nvalue > 60) are deposited continuously to horizontal
direction.
The cross section of A-site is shown in Figure 2, the
wall and struts conditions are shown in Table 1 and the
soil parameters are shown in Table 2. In this construction site, the seepage control method was adopted by
extending the earth retaining wall to the low permeable
layer Oc3, excavation width is 16.2 m, the final excavation depth is GL-21.5 m and the penetration depth
is 4.8 m.
Figure 3 shows the comparison between observed
data and design value of the earth retaining wall deflection. The observed wall deflections in east and west
sides are symmetric till the 4th step. It is confirmed that
the design value can estimates the actual phenomenon
adequately. However, since the 5th step, the observed
wall deflection of west side was larger than the east
side, which can be seen from the results of the 8th step
Earth retaining wall and each strut (A site).
Soil mixing wall (H-steel) condition

Size
(mm)
Pitch
S(m)
Length
L(m)
EI
(kNm2 /m)
Area
A(m2 )
H-588 300 12 20
0.60
27.25
399000
0.01925
Excavation condition
Strut condition
Depth
Depth
Step
(GL-m)
Stage
(GL-m)
Size
(mm)
Span
L(m)
Pitch Area
S(m) A(m2 )
0th
1st
2nd
3rd
4th
5th
6th
7th
8th
1.51
4.50
7.00
8.70
11.20
13.90
16.45
19.05
21.55
Cover beam
1st
2nd
3rd
4th
5th
6th
7th
0.51
3.50
6.00
7.70
10.20
12.90
15.45
18.05
H-588 300 12 20
H-300 300 10 15
H-300 300 10 15
H-300 300 10 15
H-300 300 10 15
H-350 350 12 19
H-350 350 12 19
H-350 350 12 19
16.25
14.76
14.86
14.66
14.76
14.46
14.46
14.46
2.00
2.50
2.50
2.50
2.50
2.50
2.50
2.50
175
0.01925
0.01048
0.01048
0.01048
0.01048
0.01549
0.01549
0.01549
in Figure 3. It can be assumed that the cause of this

phenomenon was the difference of construction process and developmental pattern of creep deformation.
Moreover, the observed data exceeded the design value
at the west side. It was considered that this disparity
occurred for the reason that the plastic zone under the
excavation bottom expanded in the Amc and Tsg layer
from the 5th step drastically. In addition, it was confirmed that the stress in the wall was controlled within
the allowable stress.
In the excavation stage at the Amc layer, the calculation result considering the 75% stress reduction
under the 5 m from the excavation bottom was shown
together in Figure 3. During an excavation in cohesive
Table 2.
Soil parameters (A site).
Soil
layer
Bottom
Internal
depth
Cohision friction
E50
2
(GL-m) N-value c (kN/m ) angle ( ) (MN/m2 )
B
Auc
Aus
Amc1
Amc2
Amc3
Alc
Tsg
Oc3
1.8
4.9
8.3
13.8
16.8
19.4
21.8
25.4
31.6
2
4
2
0
1
4
6
26
14
0
42
0
29
60
91
108
0
360
19.3
0
19.3
0
0
0
0
32.7
0
4.1
4.7
6.9
15.9
15.9
83.6
soil, if an excavation stage takes a long time, the suction of the subgrade soil will disappear due to swelling
caused by the water infiltration from continuous rainfall, which leads to reduction in strength (Hashimoto
et al., 1997). In conjunction with this arrangement,
the ultimate passive lateral pressure and coefficient
of subgrade reaction were reduced. This phenomenon
was verified by the consolidation with un-drained triaxial compression test, in which shear strength reduced
to 70% after the suction was disappear completely
and after measuring the water pressure and suction.
In short, it is proved that there is a possibility that this
phenomenon may occur (Kato et al., 2006).
In the 8th step, the calculation result considering the
stress reduction exceeded the design value which could
explain the observed data appropriately to some extent.
However, under the bottom of the excavation, especially in Tsg layer, the tendency that the design value
and calculation result exceeded the observed data. The
wall deflection distribution was different between the
observed data and design value and calculation result.
It would appear that one of the reasons for these tendencies is the deformation at the bottom of the wall
towards the excavation side.
3.2
B-site
The layer of this B-site ground constitutes the

alluvial layer, upper lower Pleistocene from the
Figure 3. Comparison between observation and design value of braced wall deflection (B site, (a) : the 4th step, (b) : the 8th
step).
176
ground surface. Especially, this construction site was

located in the Neyagawa lowland, and it is peculiar
that the very soft and sensitive alluvial clay layer
(N-value
= 0, liquid limit IL = 0.6 to 1.5, cohesion
c = 30 to 100 kN/m2 ), which is specific for the East
side of Osaka Plain, deposited with 15 to 20 m
thickness. The upper Pleistocene sandy and gravel
layer Ts & Tg and the lower Pleistocene sandy layer
Os3 (Osaka Group) constitute the second waterbearing layer under the alluvial layer.
The cross section of B-site is shown in Figure 4, the
wall and struts conditions are shown in Table 3 and the
soil parameters are shown in Table 4.
In this construction site, the seepage control method
was adopted by extending the earth retaining wall to
the low permeable layer Oc7 (about GL-42 m), too.
The bottom depth of Soil Mixing Wall (H-steel) was
extended to the Os8. The excavation width is 17.2 m,
the final excavation depth is about GL-22 m and the

penetration depth is about 5 m.
Figure 5 shows the comparison between the
observed data and design value of the earth retaining
wall deflection in the 4th and 8th excavation steps
(Oota et al., 2007). The wall deflection distribution
mode was similar both side in the 4th excavation step.
However the amount of the observed wall deflection
was two times of the design value. Moreover, in the
8th excavation step, the wall deflection distribution
mode was different in both and observed data exceed
the design value. In addition, it was confirmed that the
stress in the wall was controlled under the allowable
stress.
Table 4.
Soil parameters (B site).
Soil
layer
Bottom
Internal
depth
Cohision friction
E50
(GL-m) N-value c (kN/m2 ) angle ( ) (MN/m2 )
B
Auc
Aus
Amc1
Amc2
Amc3
Alc
Tc
Ts
Tg
Os8
0.8
2.0
4.0
8.0
13.0
16.0
19.0
20.8
23.3
26.0
39.1
2
0
2
0
0
0
3
7
42
45
60
0
27
0
42
63
76
73
129
0
0
0
19.9
0
19.9
0
0
0
0
0
37.5
38.2
41.8
Figure 4. Cross section (B site).
Table 3.
Earth retaining wall and each strut (B site).
Soil mixing wall (H-steel) condition

Size
(mm)
Pitch
S(m)
Length
L(m)
EI
(kNm2 /m)
Area
A(m2 )
H-588 300 12 20
0.60
27.52
399000
0.01925
Excavation condition
Strut condition
Depth
Depth
Step
(GL-m)
Stage
(GL-m)
Size
(mm)
Span
L(m)
Pitch Area
S(m) A(m2 )
0th
1st
2nd
3rd
4th
5th
6th
7th
8th
1.42
2.81
5.96
8.26
11.51
14.51
17.21
19.61
21.70
Cover beam
1st
2nd
3rd
4th
5th
6th
7th
0.42
1.81
4.96
7.26
10.51
13.51
16.21
18.61
H-488 300 11 18
H-300 300 10 15
H-350 350 12 19
H-350 350 12 19
H-350 350 12 19
H-350 350 12 19
H-400 400 13 21
H-400 400 13 21
17.15
16.55
16.45
16.45
16.45
16.45
16.35
16.35
2.00
2.59
2.59
2.59
2.59
2.59
2.59
2.59
177
0.01592
0.01048
0.01549
0.01549
0.01549
0.01549
0.01977
0.01977
2.2
2.2
5.5
7.4
5.6
Figure 5. Comparison between observation and design value of braced wall deflection (B site, (a) : 4th step, (b) : 8th step).
As the ground condition under the bottom of

the excavation is considered as the plastic zone in
the calculation using JSCE model, the bottom of the
wall deformed towards the excavation side in a larger
value and the wall deflection distribution had different phenomenon compared to the observation. In the
excavation stage at the Amc layer, the calculation considering the 75% stress reduction under the 5 m from
the excavation bottom was shown together in Figure 5.
Unlike the comparison result in A-site, this calculation was similar to the design because wall deflections
around the bottom of the excavation are in the plastic zone. It was impossible to explain the observed
phenomenon adequately used by some calculation
model.
The horizontal coefficient of subgrade reaction kh of
clayey ground for excavation side in the OMTB model
is determined by equation (2). This setting method of
kh was the empirical equation based on many observed
data in the case that the wall deflection was about 3 cm
(Yanagida et al., 1981). This reference bring up the
problem that kh is tend to decrease due to the increase
of the wall deflection.
In the actual construction site, as kh is depended
on the ground mechanical characteristics and some
boundary conditions and so on, it is known that

kh changes every second due to braced excavation.
For example, kh is determined as solid line by the
wall deflection function taking into consideration the
nonlinearity (Japan Road Association, 1986).
The inverse analysis based on some earth retaining
monitoring results was carried out to estimate the value
of kh . Modified Pawell Method was employed for the
inverse analysis. It is possible to obtain the optimized
solution stably on the many unknown parameter problem (Kishio, et al., 1995). Input values for the inverse
analysis are earth retaining wall deflection (angle of
inclination) and axial force of struts, and output values
are lateral pressure on the earth retaining wall and kh .
Figure 6 shows the inverse analysis results based
on the monitoring data in Osaka Subway Line No.8
touched to the Kishio, et al., 1997. The vertical scale is
the ratio of the estimation value kh by the inverse analysis to design value of kh0 . In other words, kh /kh0 = 1
means the inverse analysis results and design value are
the same.
In the case that the wall deflection was about 1 cm,
the relation between kh and kh0 was about the same
in both past actual results and Line No.8 results. In
short, the applicability of kh in the design is reasonable.
178
2.
3.
Figure 6. Dependence for brace wall deformation of kh

(touch in Kishio, et al., 1997).
4.
However in the case that the wall deflection was about
2 to 4 cm, inverse analysis results kh are smaller than
the design value kh0 , the ration kh /kh0 decreased to
about 0.5.
The relation of = 1/40 was presumed on the
assumption that kh decrease due to the increase of
the wall deflection. Figure 5 shows the recalculation
results under this relation.
In the case of the 4th excavation step, the relation
between observation and recalculation was in good
agreement. In the case of the 8th excavation step, the
wall deflection distribution mode between observation
and recalculation was still different, but the maximum
amount of wall deflection was similar. It is believed
that the cause of differences in the wall deflection
distribution mode is the limit explained by the beam
spring model in design.
In accordance with these estimations results, it is
preferable to consider the kh setting method carefully
as equation (3) with considering the traditional design
idea in the case of earth retaining design in the soft and
sensitive clayey layer, which N-value is about 0 to 2,
with thick layer (about 10 to 20 m).
CONCLUSIONS
The results are shown as follows;

1. At the A-site, the observed wall deflection in the
east and west side are symmetric till the 4th step,
and it is confirmed that the design value estimates the actual phenomenon adequately. However
5.
since the 5th step, the observed data exceeded

the design value. It was assumed that the plastic
zone expanded drastically to the penetration area
in design.
It was possible that the calculation result considering the 75% stress reduction under the 5 m from
the bottom of the excavation explained the observed
data to some extent. However, under the bottom of
the excavation, the tendency that the design value
and calculation result exceeded the observed data.
It was considered that one of the reasons for these
tendencies is the deformation at the bottom of wall
towards the excavation side.
At the B-site, the wall deflection distribution was
similar between the observation and design in the
4th excavation step. However the observed wall
deflection is two times of the design. In the 8th step,
the wall deflection distribution mode was different
in both, and observed wall deflection exceeded the
design value.
Owing that = 1/40 was presumed on the assumption that kh decrease due to the increase of the wall
deflection, the relation between observation and
recalculation was in good agreement in the case
of the 4th excavation step. In the 8th excavation
step, the wall deflection distribution mode in both
was still different, but the maximum wall deflection
was close.
It is recommended that the kh setting method carefully as equation (3) with considering the traditional
design idea in the case of earth retaining design
in the soft and sensitive clayey layer, which Nvalue is about 0 to 2, with thick layer (about 10
to 20 m).
REFERENCES
Hashimoto, T., J. Nagaya, J. Kishio and T. Shiotani : Investigation of Strength Degrading due to Swelling of the Ground
in Excavation, Proc. of the Int. Conf. on Foundation
Failures, pp.393398, 1997.
Japan Road Association: Design Guideline for Underground
Multipurpose duct, 1986 (in Japanese).
JSCE : Standard Specifications for Tunneling-2006, Cut and
Cover Tunnels, pp.142181, 2006 (in Japanese).
Kato, S., T. Konda and H. Shinkai: Effect of Suction Reduction Caused by Wetting Process on Shear Strength Characteristics Under Low Confining Pressure, JSCE Journals
C, Vol.62, No.2, pp.471487, 2006 (in Japanese).
Kishio, T., N. Nakai, H. Arimoto, T. Konda and K. Takami :
Inverse analysis example of Braced Wall used by Modified
Pawell Method, Proc. of the 50th JSCE Annual Meeting,
III-520, pp.10401041, 1995 (in Japanese).
Kishio, T., H. Oota, T. Hashimoto, T. Konda, E. Saito
and N. Kobayashi : Estimation of Lateral Pressure and
Coefficient of Subgrade Reaction during Excavation Work
in Osaka, JSCE Journals, No.560, VI-34, pp.107116,
1997 (in Japanese).
179
Kishio, T., H. Oota, T. Hashimoto and T. Konda : Some

Aspects of Designing Earth Pressures for Braced Wall
under the Bottom of Excavation, Tsuchi-to-Kiso JGS,
Vol.45, No.10, pp.2022, 1997 (in Japanese).
Oota, H., H. Ito, T. Yanagawa, T. Konda and T. Hashimoto:
Consideration of Design Method for Bracing Excavation
Based on Monitoring Results, Proc. of the 42nd JNCSFE,
2007 (in Japanese).
Osaka Municipal Transportation Bureau : Design Guideline

for temporary structure (draft), 1993 (in Japanese).
Yanagida, S., T. Watanabe, I. Yamaguchi, H. Nakamura and
S. Mizutani : The Study of Lateral Earth Pressure for Earth
Retaining Design (Part II), Proc. of the 16th JNCSFE,
14491452, 1981 (in Japanese).
180
Ground movements in station excavations of Bangkok first MRT

N. Phienwej
Asian Institute of Technology, Bangkok, Thailand
ABSTRACT: The characteristics of movements of diaphragm wall and ground in the excavation of 18 stations
of the first Bangkok underground MRT line were evaluated. Three modes of deflected shapes of the walls were
observed at different excavation depths, namely cantilever mode, braced modes with bulge in soft clay and bulge
in stiff clay. The ratio of maximum lateral wall deflection developing with excavation depth and the ratio of
ground surface settlement to excavation and the normalized zone of ground surface settlement varied with the
mode of wall deflection. Back-calculation of undrained soil moduli for different soil layers were made from wall
movement data of three selected stations using the 2-D linear elasto-plastic FEM analyses. The modulus values,
which were higher than those commonly obtained from conventional triaxial tests, can be used as guideline for
future works in Bangkok.
INTRODUCTION
Deep excavation by means of strutted concrete

diaphragm walls is often used for construction of
multi-level building basements in Bangkok soft soil.
It is superior to the excavation with steel sheet piles
for control of ground movement to avoid damages to
adjacent structures. Prior to the construction of the first
Bangkok MRT project, few studies were made on the
characteristics of the ground movement and its prediction (e.g. Phienwej and Gan, 2003 and Teparaksa
et al, 1999. etc.). However, it was the implementation of the first Bangkok MRT subway line that
provided systematic and comprehensive monitoring
data on the excavation of station boxes that allowed indepth evaluation on the characteristics of the wall and
ground movements associated with deep excavation in
Bangkok subsoil conditions using strutted diaphragm
walls. The project involved the deepest ever excavation
made in Bangkok to date. The excavations of all station
boxes were fully instrumented. Monitoring data from
18 station excavations were compiled, summarized and
interpreted.
The construction of the first MRT underground
project in Bangkok, the Mass Rapid Transit Initial System Project (MRT ISP Blue Line) was started in 1998.
Prior to that there were public doubts on technical viability of the construction and operation safety of underground MRT in Bangkok soft soil. That pessimistic
outlook led to a call for an in-depth investigation on the
application and performance of the excavation method
and support systems to be integrated in the excavation
in soft and subsiding Bangkok ground. The contracts

made its compulsory that full instrumentation program
be implemented during excavation for design verification and safety assurance. Evaluation on the actual
performance at sites was performed to confirm the sufficiency in the design of the support systems for the
MRT station excavations. A comprehensive study was
made on the aspect of the wall and ground movements
(Hooi, 2003) and the salient points from the study are
reported herein.
PROJECT DESCRIPTION
The ISP Blue Line is the first underground MRT line

constructed in Bangkok. It comprises 22-km-long twin
single-track tunnels, 18 stations and a depot. The horizontal alignment mainly follows the right of way of city
roads. Construction of the underground structures was
implemented under two fast track design-built contracts, each having approximately the same amounts
of work. The South Contract involved construction of
a twin bored tunnels from the inter-city railway terminal at Hua Lamphong eastwards for 5 km beneath the
busy Rama IV road to the Queen Sirikit National Convention Center, then 4.5 km north beneath the narrow
business Asoke road, and Ratchadaphisek road ending
on the surface near the depot. Works of the North Contract continued for 4.5 km north along Ratchadaphisek
road to Lad Phrao road then turned west to Chatuchak
Park and finally terminated beneath the Bang Sue
181
by approximately 23 m from the original hydrostatic

profile. The groundwater pumping has also created
regional subsidence throughout Bangkok metropolis.
At the locations of the project, a perched water table
is typically encountered in Made Ground. Below this
horizon, hydrostatic conditions are generally found to
a depth of approximately 8 m to 10 m depending on
the location and thickness of First and Second Sand
layers. Typically, the upper few metres of sand layers
underlying First Stiff Clay and Very Stiff Clay have
been dewatered.
railway station. The 18 stations of the project are as

follows:
South Contract
North Contract
1. Hua Lamphong Station

2. Sam Yan Station
3. Silom Station
4. Lumphini Station
5. Bon Kai Station
6. Sirikit Centre Station
7. Sukhumvit Station
8. Phetchaburi Station
9. Rama IX Station
1. Thiam Ruam Mit Station

2. Pracharat Bumphen Station
3. Sutthisan Station
4. Ratchada Station
5. Lad Phrao Station
6. Phahonyothin Station
7. Mo Chit Station
8. Kamphaeng Phet Station
9. Bang Su Station
STATION CONSTRUCTION
The MRT stations had following features:

3
GROUND CONDITIONS
Bangkok is situated on the southern part of the low

lying Chao Phraya plain, which extends north from the
coast line at the Gulf of Thailand up to approximately
350 km and spans east-westward up to 150 km. The
flat topography plain is covered with a thick marine
clay layer, which overlies a very thick series of alluvial deposits of alternating stiff to hard clay and dense
to very dense sand to gravel. The thick soft clay layer
generally extends from the ground surface to a depth of
12 to 15 m. The soft clay which is known as Bangkok
soft clay has high water content (70120%), high
plasticity, low strength and high compressibility.
The shallow subsoil of the upper 35 m zone is relatively uniform and generally consists of layers of soft
to medium clay, stiff to hard clay and sand. Below this
shallow zone, alternating layers of stiff to hard clay and
dense sand exist to a great depth. Bedrock is found at
depths more than 450 m. The typical subsoil profile
for the first 50 m depth is listed as follows:
Made Ground
Bangkok Soft Clay
First Stiff Clay
Medium Dense Clayey Sand, Sandy Clay and Silty
Clay
Very Stiff Sandy Clay/First Bangkok Sand
Second Hard Clay
Second Bangkok Sand
Maconochie (2001) summarized the general soil
profile and properties at the Bangkok MRT ISP Blue
Line. The variation in soil profile along the alignment was observed primarily in the Very Stiff Sandy
Clay/First Bangkok Sand layer and the soils immediately below it. Figure 1 show the soil profile along the
MRT alignment and at the stations.
Deep well pumping in Bangkok and its environs
has reduced the pore water pressures in the sand layers
Typically, three levels of structure, with a centre platform that is fed by stairs and escalators between two
lines of columns down the middle of the station.
Up to 230 m long and approximately 25 m wide,
excavated up to a depth of 25 m to 30 m below the
ground surface.
The perimeters were of diaphragm walls, 1.0 m thick
and 30 m to 35 m deep and solid in-situ reinforced
concrete slabs, typically 0.9 m thick for the roof
slabs, 0.7 m thick for the intermediate slabs and
1.75 m thick for the base slabs, which were used
as the excavation support system and permanent
structures later on.
There were three stacked stations, at Samyan, Silom
and Lumphini Stations, due to space constraints
caused by the existence of the foundation piles of the
long road flyover and a water transmission tunnel at
these busy intersections.
There was a side platform station at Bang Su Station, with two levels only, to accommodate the track
alignment for future elevated extension of the line
to the north.
The remaining stations were constructed as Centre
Island Platform stations.
Pracharat Bamphen and Sutthisan Stations incorporated intersection road underpasses on the roofs of
the stations.
Silom, Lad Phrao and Phahonyothin Stations were
excavated underneath foundation piles of existing
road flyovers and thus, the station structures and
foundation were designed to support the flyovers
via cross-beam and underpinning bored piles.
The top down construction technique was adopted
for all station box excavations with diaphragm walls
and concrete slabs as the excavation support system,
which was later utilized as the permanent structures
of the stations. The designs were made with an aid
of FEM analyses. The excavation depths and the toe
depth of the diaphragm walls of all station excavations
182
Figure 1. Soil profile along MRT alignment and at the stations.
are summarized in Fig. 2. It should be noted that the

ratio of the depth of the wall embedment to the depth
of excavation was significantly different between the
two contracts, primarily due to the difference in the
design criteria adopted by the two different designers.
INSTRUMENTATION DATA
The measurement data from inclinometers and surface settlement points were compiled and interpreted
to evaluate the overall performance of the station excavations in Bangkok subsoil using diaphragm walls.
The data were screened to preclude movements not

expressly related to the excavation and support installation, such as diaphragm wall construction, temporary decking works and the initial 2.3 m excavation that
involved driving sheet piles, backfilling and extracting the sheet piles subsequently. A detailed study of the
instrumentation data obtained was undertaken by comparing observed ground movement among the station
excavations. Factors that may result in such patterns
of data was examined and deduced, such as:
Station box configuration and dimension
Construction sequences
183
NORTH CONTRACT
0.0
Maximum Lateral Wall Movement/Excavation Depth, dHmax/H

0.2
0.4
0.6
0.8
1.0
1.2
1.4
Excavation Depth, H (m)
0
5
10
15
20
25
Cantilever mode (H=1.6m-4.0m)
Braced mode-soft soil (H=6.5m-10.8m)
Braced mode-stiff soil (H=12.4m-16.7m)
SOUTH CONTRACT
Figure 2. Depth of excavation and toe depth of D-wall.
0.0
Maximum Lateral Wall Movement/Excavation Depth, dHmax/H

0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
1.8
Excavation Depth, H (m)
0
5
10
15
20
25
30
35
Cantilever mode (H=1.6m-4.0m)
Braced mode-soft soil (H=6.5m-10.8m)
Figure 4. Normalized maximum wall deflection versus

excavation depth.
Figure 3. Modes of wall deflection.
Variation in soil profile and properties

Temporary works or presence of structures adjacent
to excavation box
5.1
Lateral wall movement
In general, most inclinometers all station excavations

showed that the cantilever mode was the most predominant of wall deflection shape at the initial excavation
stage, while the braced excavation mode developed in
the subsequent stages as the excavations were deepened. Figure 3 shows the three modes of deflected
shape of wall movement, which occurred at different excavation depths. The cantilever mode was most
common during the first excavation stage. The braced
excavation mode with bulge in soft soil prevailed at the
second and third excavation stages. This mode continued to dominate the pattern of lateral movements
for North Contract, but data of South Contract exhibit
that the braced mode with bulge in the underlying
stiff soil layers was more predominant. This different behavior may be attributed to the difference in soil
profiles. In addition, South Contract had three deep
station excavations of vertically stacked platforms.
The maximum lateral wall deflection of the 18 station excavations were in the range of 1047 mm. The
monitoring data also showed that there were significant variations in the shape and magnitude of the
lateral wall movement within some of the station excavations. The variation may be attributed to a number
of factors including the variation in soil profile and
properties over the plan area of the station, adjacent
temporary surface works, and confinement from road
pavements and buried utility structures. The area of
the excavation was quite large (about 25 m wide and
200 m long). For the three stations adopted for the
detailed analysis in this study, the variations in the
maximum lateral wall movements were as follow: 28
38 mm for the deepest Silom Station, 15 to 40 mm
for Sirikit Station, and 1827 for Thiam Ruam Mit
Station. Figure 4 summarizes the ranges of the normalized maximum lateral wall movement with excavation
depth (Hmax /H) versus excavation depths recorded by
all inclinometers at the 18 stations. The plots show
that in the first stage cantilever mode excavation the
value of Hmax /H was as high as 1.60. The maximum value of Hmax /H decreased with the increase
in the excavation depth in the braced modes of the
184
For shallow excavation under the cantilever mode the

value of vmax /H can be higher. The characteristics
of ground surface settlement behind the excavation
with diaphragm walls can be categorized according to
the mode of wall deflection of excavation depth. The
zones of ground settlement for the three modes of wall
deflection in Bangkok soil are marked in figure. In
addition, the three categories of ground movement in
braced wall excavation for flexible walls (sheet piles or
soldier piles) suggested by Peck (1969) are also shown
in the plot. The level of ground settlement in excavation with diaphragm wall in Bangkok soft soil is much
smaller than that predicted by Pecks chart for flexible
wall. However, the influence zones of ground settlement were wider than those suggested by Pecks. This
finding can be used as a general guideline for prediction ground surface settlement from deep excavation
with diaphragm walls in Bangkok subsoil condition.
Figure 5. Normalized settlement versus distance from wall.
walls. The decrease in the value with depth was due to

the change in soil condition at the excavation bottom
from soft clay to stiff clays as the excavation depth
increased. When the excavation depth reached stiffer
soils, Hmax /H decreased and the range of the value
became narrow. When the excavation bottom was still
in the soft clay layer, the maximum value of Hmax /H
was smaller than 0.5. The value was smaller than 0.2
when the excavations were deeper, in stiff clays.
5.2 Ground surface settlement
The maximum ground surface settlement was
observed at 58 mm at Bon Kai Station and 75 mm at
Pracharat Bumphen, in South and North Contracts,
respectively. Figure 5 shows plot of the ratio of maximum settlement to excavation depth (vmax /H) versus
the distance from excavation normalized by the excavation depth (D/H). For shallow excavations under
cantilever mode of wall deflection, the zone of ground
settlement may extend up to D/H of 7 to 10. As for
the excavation depths while the wall deflection was
under the braced mode with bulge in soft clay layer,
the vmax /H may extend up to D/H of 7. For deeper
excavations while the wall was deflected under the
mode of bulge in stiff clay layer, the zone of ground
settlement may extend up to a distance of D/H of 4
and vmax /H value may be as high as 0.35 Under the
braced mode in soft clay, vmax /H value may reach 0.55.
PREDICTION OF MOVEMENTS
For the design of the MRT station excavations of

both contracts, FEM analysis were made to determine ground movement and forces on the diaphragm
walls and bracings. The analyses mainly utilized linear elasto-plastic Mohr-Coulomb soil parameters. The
parameters were obtained from the soil investigation
and testing program made for each station excavation.
Triaxial compression tests as well as pressuremeter
tests were conducted to determine the values of soil
modulus for design analysis. The instrumentation data
provided valuable information to evaluate the appropriateness of the soil model and the soil parameters
used in the design calculation. In this study, monitoring data from three representative station excavations,
i.e. Silom, Sirikit and Thiam Ruam Mit Stations,
were examined in details and suitable soil parameters were back-calculated using a continuum FEM
analysis. Computer code PLAXIS 2D was adopted in
the study. Effective stress strength parameters were
adopted in the undrained analysis. The drawdown phenomenon of the piezometric levels was considered in
the simulation.
Silom Station was the deepest excavation in
Bangkok to date (32.6 m deep), with a vertically
stacked platforms thus it had four levels of slab below
the roof. The station was designed to underpin the
existing flyover roadway running over the station
length. A dense sand layer of the first Bangkok Sand
was encountered from depth 8.5 m above the final
excavation level. Hence, the excavation required dewatering. The diaphragm wall was toed into the Second
Sand layer.
Sirikit Station had the typical configuration of the
centre island platform with three levels below the roof
slab. The first stage and final excavation depths of
185
3.65 m and 23.6 m respectively, which were similar to

majority of other stations.
Thiam Ruam Mit Station was selected because the
soil profile was slightly different from those found in
the first two stations. The area had a thicker First Stiff
Clay layer with lenses of clayey sands. In addition, the
first stage excavation was very shallow with roof slab
was only at 1.8 m depth.
The back-calculation using the lateral ground movement data from the excavations of the three stations
suggested the suitable undrained soil modulus parameters as follows.
Soft and Medium Clay : Eu = 500 Cu kN/m2
First Stiff Clay
: Eu = 700 N60 kN/m2
Clayey Sand and Silty/ : Eu = 900 N60 kN/m2
Sandy Clay
Second Hard Clay
: Eu = 1600 N60 kN/m2
Third Hard Clay
: Eu = 2500 N60 kN/m2
where Cu is the corrected field vane shear strength and
N60 is the corrected SPT N value according to Liao
and Whitman (1986).
These back-calculated values of soil modulus are
higher than those commonly obtained from conventional laboratory triaxial tests. It reflects the modulus
values at low strain level which would be the dominating response of soil in the excavation problem (Mair,
1993). The soils would be mainly under unloading
condition of stresses.
In similar early studies, Phienwej and Gan (2003)
and Teparaksa (1999) both proposed the same modulus parameter of the soft clay as Eu = 500cu . While
for stiff clay, the value of Eu = 1200Cu and 2000Cu
were suggested, respectively. Based on the relationship of Cu = 0.6N60 kN/m2 typically used for Bangkok
subsoil, the parameters are equivalent to Eu = 720N60
and 1200N60 kN/m2 , respectively. The back-calculated
values from this study were comparable to those
suggested by Phienwej and Gan (2003).
CONCLUSIONS
The following conclusions can be drawn from the study

of the wall and ground movements in the excavation of
the stations of the first Bangkok MRT underground.
Three modes of deflected shapes of the wall movement were observed at different ranges of excavation depth. Mode 1: Cantilever mode (H = 1.6
m4.0 m), Mode 2: Braced mode with bulge in soft
clay layer (H = 6.5 m11 m), and Mode 3: Braced
mode with bulge in stiff soil (H = 12.4 m 32.6m).
The maximum lateral wall movement (Hmax ) was

smaller than 47 mm in both contracts. The normalized wall deflection, Hmax /H in the cantilever mode
of movement was as high as 1.60, while it was
reduced to no more than 0.60 and 0.40 in the latter
stages of excavation when the wall deflection developed in the braced mode with bulge in soft clay and
braced mode with bulge in stiff soil, respectively.
The maximum ground surface settlement (Vmax )
was 58 mm for South Contract and 75 mm in North
Contract. The normalized maximum ground settlement with excavation depth, vmax /H, was smaller
than 0.55 and 0.35 for Modes 2 and Mode 3 wall
deflection, respectively. The normalized distance
from excavation of the zone of ground settlement,
D/H, varied from 7.0 and 4.0 for the two modes
of wall deflection. In the initial excavation stage
of cantilever mode, the values of both normalized
settlement and distance of ground movement were
higher that those in the braced modes.
Back-calculation of soil moduli of different soil
layers using monitoring data from three selected
stations showed higher values than those commonly obtained for conventional laboratory tests.
The values are: Soft and Medium Clay: Eu = 500cu ,
First Stiff Clay Eu = 700 N60 kN/m2 , Clayey Sand
and Silty/Sandy Clay Eu = 900 N60 kN/m2 , Second
Hard Clay Eu = 1600 N60 kN/m2 , Third Hard Clay
Eu = 2500 N60 kN/m2
REFERENCES
Hooi, K.Y. 2003. Ground Movements Associated with Station
Excavations of the First Bangkok MRT Subway. Master
Thesis, Asian Institute of Technology, Bangkok.
Liao, S. & Whitman, R.V. 1986. Overburden Correction Factor for SPT in Sand. Journal of Geotechnical Engineering.
American Society of Civil Engineers 112(3): 373377.
Mcconochie, D. 2001. Geotechnical Completion Report
MRTA Chaloem Ratchamongkhon Line. A CSC Report
submitted to the MRTA.
Mair, R. 1993. Developments in geotechnical engineering
research: application to tunnels and deep excavations.
Unwin Memorial Lecture 1992, Proceedings of Institution
of Civil Engineers, Civil Engineering, 93, Feb: 2741
Phienwej, N. & Gan, C.H. 2003. Characteristics of
Ground Movements in Deep Excavations with Concrete
Diaphragm Walls in Bangkok Soils and their Prediction. Journal of The Southeast Asian Geotechnical Society
34(3): 167175.
Teparaksa, W., Thasnanipan, N. & Tanseng, P. 1999. Analysis
of Lateral Wall Movement for Deep Braced Excavation in
Bangkok. Proc. of AIT 40th Anniversary Conference, AIT,
Bangkok, Thailand.
186
Numerical modelling and experimental measurements for a retaining wall

of a deep excavation in Bucharest, Romania
H. Popa, A. Marcu & L. Batali
Technical University of Civil Engineering, Geotechnical and Foundations Department, Romania
ABSTRACT: Although civil engineers dispose of various calculation methods for retaining structures, none
of them have definitely imposed itself, each one bringing its own benefits or limitations. Finite Element Method
(FEM) offers the benefits of complex models allowing taking into account the majority of soil structure
characteristic parameters. However, the experience shows that the differences between the experimental and
the calculation results are often quite important. The paper presents the case history of a diaphragm wall for
a deep basement of a new building in centre of Bucharest. Nearby the new building there is an ancient cathedral historic monument. The deep excavation is also neighboring at different distances with another existing
buildings and a heavy trafficked road. All these conditions led to choose the top-down technology in execution of the basement. The numerical results obtained by FEM are compared with the measurements recorded
during the construction. The differences between the obtained values (displacements) are comprised between
15% and 75%, depending on the enclosure sides. The main factors leading to these differences are the soil
parameters.
2 WORK AND SITE DESCRIPTION
INTRODUCTION
The effect of deep excavations on neighboring structures can become important and therefore special
measures have to be taken in the design and monitoring
of retaining walls.
Calculation of such structures must be based on
methods taking into account the soil structure interaction and, with this respect good soil knowledge is
indispensable.
Paper presents the case of a diaphragm wall for
a deep basement in the very centre of Bucharest.
The basement is developed on 4 underground levels and needs an excavation of about 15 m deep. The
groundwater level is at about 6 m depth. The ground
is composed of alluviums layers comprising medium
soft silty clays, as well as fine to coarse medium
dense sand. Near the new construction, at about 6 m
distance, there is an ancient cathedral, classified as
historical monument. As well, the pit has on another
side some buildings, while on the other two sides it
is delimited by a heavy traffic road. All these conditions led to choose the top down technology for
building the infrastructure, in order to have minimum
deformations and displacements of the wall, so that
the integrity of the neighboring buildings not being
affected.
2.1
Site and geometrical characteristics
The new building is located at the centre of Bucharest,

next to the Romano Catholic Cathedral St. Joseph.
Figure 1 presents a photo of the site.
As it can be seen on the photo, on the Western side
the pit is very close to the St. Joseph Cathedral (about
Figure 1. Location of the deep excavation.
187
Table 1.
Design values for the geotechnical parameters.
Stratum
Clay (1)
Sand & gravel
Clay (2)
Silty clayey
sand
Clay (3)
Fine sand
K0
Thickness,
E,
c
m
kN/m3 MPa ( ) kPa
4.00
5.50
7.50
2.00
19
20
20
20
25
40
50
75
25
38
22
28
30
0
50
20
0.7
0.4
0.7
0.5
4.00
7.00
20
20
75
75
22
36
50
0
0.4
0.4
The ground lithology and the design values of the

geotechnical parameters are shown in table 1 (level
0.00 m represents ground level).
The groundwater level is +74.0 m and, according
to the site investigations, it can vary with 1.00 m.
A second aquifer, confined, has been found between
+63.0 +61.0 m (within the sandy layer).
3
3.1
Figure 2. Cross section through the diaphragm wall.
6 m distance), on the Southern side there is a building

complex at about 10 m distance, while on the two other
sides there is a public road with heavy traffic.
Figure 2 shows a cross section through the
buildings infrastructure, presenting also the ground
lithology.
The diaphragm wall is made of panels 80 cm thick
and 21 m deep. The embedment depth has been established considering the wall stability and excavation
bottom imperviousness. Dewatering has been performed only inside the enclosure, the groundwater
level outside being left unchanged to avoid undesirable settlements of the ground around the pit due to
dewatering, which would be in addition to the inherent settlements due to excavation inside the enclosure
and to the erection of the new structure, which could
affect especially St. Joseph Cathedral.
2.2
Numerical model
Numerical modeling has been performed using 2-D

FEM, the model having 1933 elements and 5866
nodes.
For the soil, a perfect elasto-plastic constitutive law
has been used, with Mohr Coulomb criteria, using the
geotechnical parameters issued from laboratory and
in situ tests.
3.2
Calculation stages
Calculations have been organized in 6 stages, following the technological phases:

phase 0 initialization of the stress state in the
ground;
phase 1 excavation down to the lower level of the
first floor slab; execution of the first floor slab;
phase 2 excavation below the first floor slab down
to the lower level of the second floor slab and dewatering inside the enclosure; execution of the second
floor slab;
phase 3 excavation between floor slabs no. 2
and 3; execution of floor slab no. 3;
phase 4 excavation below the third floor slab down
to the final level (15.00 m).
Geotechnical characteristics of the ground
The geotechnical parameters are specific for Bucharest

area, which is characterized by alluvium soils. The triaxial tests performed with imposed stress path allow a
direct determination of the shear strength parameters
(, c), of the secant modulus (E) and of the earth coefficient at rest (k0 ) for the clayey layers; for the sands
these parameters have been established based on SPT
tests.
NUMERICAL MODELLING
3.3
Results
Figure 3 presents the evolution of the horizontal displacement of the wall as a function of the execution
stages.
188
Figure 3. Horizontal displacements function of calculation

phases.
Figure 5. Shear force function of calculation phases.

Table 2.
Floor slab reaction forces.

Force, kN/ml
Phase 1
Phase 2
Phase 3
Phase 4
Floor 1
Floor 2
Floor 3
142.2
2.3
6.7
456.5
249.2
729.2
According to the calculations, the maximum horizontal displacements of the wall are of about 15 mm,
at 15 m depth. Due to the top-down technology, the
shape of the displacement curves shows greater values
in the lower part of the wall. The displacements of the
upper part are practically blocked by the already built
floor slabs.
The upper maximum horizontal displacements are
estimated to be of 5 6 mm.
Bending moment and shear force graphs are shown
figures 4 and 5, respectively.
The reaction forces on the basements floor slabs
are shown table 2.
4
Figure 4. Bending moment function of calculation phases.
MEASUREMENTS
In order to record the influence the deep excavation

and, moreover the whole new building, has on the
189
Figure 6. Experimental measurements.
neighbouring structures, a monitor of the displacements has been performed.

The retaining wall was equipped on each side with
inclinometers for measuring the horizontal displacements and, thus, the walls deformation.
Marks were installed on the surrounding buildings
to monitor their settlements. As well, extensometers
were installed, just behind the wall for measuring
grounds settlements.
Another monitoring concerned the groundwater
level outside the enclosure. For this purpose, wells
were drilled and equipped as piezometers, located on
each side of the enclosure. The measurements showed

that the dewatering works inside the enclosure didnt
affect the groundwater level outside.
Figure 6 shows the monitoring equipment used for
the St. Joseph Cathedral side.
It can be seen that, in order to reduce furthermore
the risks of a negative influence of the retaining wall on
the Cathedral, between these two first a stabilization
wall was built using cement-based injections.
Figure 6 shows also the measured lateral displacements of the wall during the last excavation stage (stage
4). It can be seen that the shape of the displacement
190
curve corresponds to the one obtained by numerical

calculation. But the values are much less than the
estimated ones.
So, at the upper part of the wall, the maximum displacements are of about 2 mm, representing about 50%
of the estimated ones, while at the lower part the differences are more important, the measured values being
only 25% of the calculated ones.
This important difference between calculations and
measurements can be due also to the protection injection screen located on this side of the enclosure. On
the other sides, where no such protection has been
installed, the maximum walls displacements were of
about 13 mm, being quite closed to the estimated value
(85%).
Anyway, estimations are still higher than the measured values for all excavation stages.
Concerning the ground settlement behind the
retaining wall (at about 1 m distance), extensometers
showed a maximum value at ground level of about
7 mm. The settlement evolution versus the depth can
also be seen in figure 6 and one can note that it becomes
negligible at about 23 m below walls toe.
Marks fixed on the St. Joseph Cathedral indicated a
maximum settlement of 0.7 mm, its integrity not being
endangered.
From this point of view it can also be noted the
beneficial role of the injection screen, the difference
of settlements on one side and the other of the screen
being substantial.
Concerning the other neighboring buildings located
at about 10 m distance, maximum settlements of about
3 mm were recorded, insignificant for their stability.
5
CONCLUSIONS
Retaining structures imply complex soil structure

interaction phenomena. A correct estimation of their
behavior is possible only by using numerical models, allowing a complex modeling of the system
formed by the retaining wall, foundation ground and

neighbouring buildings.
Even when such methods are used, the results can
present significant differences from the real behavior. The reasons for these differences are many, among
them:
incertitude regarding the geotechnical parameters
used for the calculations, especially when complex
constitutive laws are used for the ground;
difficulty in estimation of the initial stress state in
the ground, taking into account its lithology, the
presence of neighbouring structures, the execution
of the retaining wall itself etc.;
complexity of the numerical model itself, considering all implied parameters;
three-dimensional behaviour of the retaining
structure.
In order to obtain reliable results using numerical modeling it is important to calibrate and validate
the model based on experimental measurements performed on similar structures and in similar site conditions. The experience in such modeling is also an
important aspect.
REFERENCES
Marcu, A., Popa, H., Marcu, D., Coman, M., Vasilescu, A. &
Manole, D. 2007. Impact deep excavations on neighboring
buildings, National Conference AICPS, 1 June 2007 (in
Romanian).
Marcu, A. & Popa, H. 2004. Calculations and measurements
of deformations and displacements of a retaining wall
for a deep excavation and of the neighboring structures.
10th National Conference of Soil Mechanics and Foundation Engineering, 1618 September 2004, Bucharest,
Romania, pp. 311322 (in Romanian).
Popa, H. 2002. Contributions to the study of the soil
structure interaction in case of underground structures,
PhD thesis, Technical University of Civil Engineering
Bucharest, p. 311.
191
3D finite element analysis of a deep excavation and comparison

with in situ measurements
H.F. Schweiger
Computational Geotechnics Group, Institute for Soil Mechanics and Foundation Engineering, Graz
University of Technology, Austria
F. Scharinger & R. Lftenegger

GDP Consulting Engineers, Graz, Austria
ABSTRACT: The paper describes the analysis of a deep excavation project in clayey silt in Salzburg. The
excavation was supported by a diaphragm wall, a jet grout panel and three levels of struts. Because of insufficient
information on the material properties of the jet grout panel the stiffness of it was varied in a parametric study. The
effect of taking into account the stiffness of a cracked diaphragm wall on the deformations was also investigated.
In some of the 3D calculations a non-perfect contact between diaphragm wall and strut was simulated by means
of a non-linear behaviour of the strut. The evaluation of the results and comparison with in situ measurements
showed that analyses which took into account the reduced stiffness of the diaphragm wall due to cracking achieved
the best agreement with the measurements. Furthermore settlements of buildings could be best reproduced by
the three-dimensional model.
INTRODUCTION
Soft subsoil deposits in Austria are mainly fresh water

deposits, sedimented in the post-glacial lakes after the
boulder periods. These deposits are known as lacustrine clays on the foothills of the Alps. One example
for a widespread lacustrine clay deposit is the basin
of Salzburg, where the city of Salzburg is situated
on subsoil sediments, which partly show a thickness
up to 70 m, called Salzburger Seeton, which can be
classified as clayey silt.
In the design stage of deep excavations in such problematic soils finite element calculations are a useful
tool to obtain reasonably realistic predictions of deformations expected. In practical engineering 2D-models
are still prevailing, but 3D-model become increasingly
attractive. It will be shown, and this is the main purpose of this paper, that the best overall match with in
situ measurements, in particular with respect to surface
displacements behind the wall, is achieved with 3Dmodels. If only wall deflection is considered also 2D
analyses show reasonable agreement. The mechanical
behaviour of the soil is modelled with an elasto-plastic
constitutive model, namely the Hardening Soil model
as implemented in the finite element code Plaxis
(Brinkgreve 2002). For the project the class A 2D
analysis predicted the overall deformation behaviour
with sufficient accuracy from a practical point of view,
but a more detailed comparison with in situ measurements has been made after construction involving 3D
finite element analyses. Furthermore some details with
respect to the strutting have been changed during construction which have not been taken into account in
the original analysis.
The input parameters for the constitutive model
have been determined not solely from site investigations but also from previous experience of finite
element analyses under similar conditions (see e.g.
Schweiger & Breymann 2005).
In the following a brief description of the problem
will be provided together with the material parameter used. The different assumptions with respect to
modelling the diaphragm wall and the jet grout panel
are discussed. Finally results from various 2D and 3D
analyses are compared with in situ measurements of
wall deflection and surface displacements.
2
2.1
PROBLEM DESCRIPTION AND MATERIAL

PARAMETERS
Project description
A cross section of the excavation with strut levels

and final excavation depth is shown in Figure 1. In
plan the excavation is roughly square, approximately
193
Table 1.
Stiffness parameters for soil layers.

E50
(MPa)
Soil layer
(04 m)
Clayey silt
Table 2.
Eoed
(MPa)
Eur
(MPa)
pref
(kPa)
ur
12
0.0
40
0.2
37.6
37.6
150.4
0.30
100
0.2
Strength parameters for soil layers.
Soil layer (04 m)

Clayey silt 1
c
(kPa)
( )
( )
5
30
28
26
0
0
Figure 1. Cross section of excavation and strut levels.
19 20 m, which of course must raise doubts whether

a 2D analysis is at all appropriate in this case. Attention
is paid to the fact that a jet grout panel just below the
final excavation level has been constructed to act as
lateral support. This has been constructed before the
start of the excavation and allowed excavation without
installing a fourth strut level. Groundwater lowering
inside the excavation was achieved by vacuum wells
(commonly used in Salzburg) which extended below
the excavation level in order to reduce uplift.
The construction sequence is closely reflected in
the analysis. Starting from the initial stress state
(K0 = 0.55 for all layers) and the loads of the foundations of the neighbouring buildings (80 kN/m2 for
the Novotel, 200 and 250 kN/m2 for the strip footings of Object 24) the wall and jet grout panel
have been introduced wish-in-place. Then excavation
steps, groundwater changes and installation of struts
have been modelled in a step by step analysis. Soil
behaviour below 20 m is assumed to be undrained,
above 20 m, due to the presence of thin sandy layers,
as drained.
2.2
Material parameters
The soil parameters used in the analysis for the top

soil layer (04 m below surface) and the clayey silt
are summarized in Tables 1 and 2. As mentioned previously, parameter determination is not only based on
site investigations and laboratory experiments but also
from experience of back analyses of other deep excavations in Salzburg. Therfeore soil parameters have
not been varied in this study. In Table 1 E50 , Eoed
and Eur are the reference stiffness in primary loading (for deviatoric and oedometric stress paths) and
unloading/reloading respectively.
The axial stiffness of the struts (Table 3) differs for
the three levels, the material behaviour is assumed to
Table 3. Axial stiffness of struts.
Strut level 1
Strut level 2
Strut level 3
Table 4.
tions.
EA
(kN)
spacing
(m)
3.234E6
1.067E7
5.334E6
3
3
3
Parameters for wall, jet grout panel and founda-
Diaphragm wall
Jet grout panel
Foundations
E
(kN/m2 )
Rinter
UCS
(N/mm2 )
2.9E7
5.0E5
3.0E7
0.2
0.2
0.2
0.7
0.7
0.7
18.8
2.25
be linear elastic. Table 4 lists the basic set of parameters used for diaphragm wall, jet grout panel and
the foundation structures of Novotel and Object 24.
In the 2D analyses a Mohr-Coulomb failure criterion
has been used for wall and jet grout panel whereas the
cohesion was chosen in such a way to obtain the uniaxial compressive strength (UCS) as listed in Table 4,
assuming = 45 . Tension cut-off was set to UCS/10.
Rinter denotes the reduction of soil strength to model
wall friction. In the 3D analyses the wall was elastic and stiffness was either assumed to correspond to
uncracked conditions or cracked conditions. The
stiffness properties of the jet grout panel have been varied because of the significant uncertainty in obtaining
reliable values for the in situ stiffness of such panels.
194
Figure 3. 2D finite element mesh.
Figure 2. Location of points used for comparison.
IN SITU MEASUREMENT PROGRAMME
The existence of structures in the close vicinity of the

excavation required a careful observation of deformations during construction. Therefore, about 30 settlement gauges were installed to monitor the settlements
outside the excavation, in particular of the adjacent buildings. In addition, four inclinometers were
installed in the diaphragm walls in order to measure
the horizontal deflection of the wall in all construction
stages. Two of them were located approximately in the
cross section chosen for the 2D analysis, i.e. along
the centre line of the excavation. Figure 2 depicts the
points chosen for the comparison of measurement and
analysis for settlements.
NUMERICAL MODELS
As mentioned previously 2D and 3D analyses have

been performed using Plaxis 2D and Plaxis 3D Foundations.The 2D model consists of approximately 2,300
15-noded elements (Figure 3) and the 3D model of
approximately 11,000 15-noded wedge elements (Figure 4). Lateral boundaries are fixed in horizontal
direction and the bottom boundary in vertical and
horizontal direction in both models. It can be seen
that the 3D mesh is much coarser as compared to
the 2D mesh but studies performed on the 3D model
showed that a mesh with more than 20,000 elements
resulted in only marginal differences in displacements.
However, bending moments are more sensitive to discretisation and a stability analysis would certainly not
yield correct results with the mesh adopted for the 3D
analyses.
RESULTS OF 2D MODEL
Four different analyses have been performed with the

2D model:
Variation 1 (V1): Wall and jet grout panel elastic
with elastic properties according to Table 4.
Figure 4. 3D finite element mesh.
Variation 2 (V2): Diaphragm wall modelled as

elastic-perfectly plastic material with UCS as given
in Table 4.
Variation 3 (V3): V2 and increase of stiffness of jet
grout panel by a factor of 3.
Variation 4 (V4): V3 and increase of tension cut-off
in diaphragm wall by a factor of 2.
Figures 5 and 6 compare the deflection of the
wall for the final construction stage for all four
analyses with the measurements obtained from the
inclinometers.
It follows that the different assumptions made have
little influence on the results in the upper part of the
wall because in this part the deformations are governed
by the struts. Results for the right wall compare well
with measurements in the upper part, for the left wall
this is not the case. For the lower part only V3 and V4
produce a reasonably match and it turned out that it
is difficult to obtain the wall curvature as measured at
the location of the jet grout panel.
Figures 7 to 9 show a comparison of calculated and
measured vertical displacements at various points on
the ground surface. The two sets of squares in each diagram represent pairs of settlement gauges which are in
close distance to the points picked from the numerical
analysis at various stages of construction (the dates are
given within the diagram, the axis represents calculation steps, representing the progress of construction
195
Figure 5. 2D analysis: wall deflection left wall.

Figure 6. 2D analysis: wall deflection right wall.
with time). Only for point H a reasonable agreement between calculation and measurements could be
achieved, although one has to mention that absolute
values are very small, with about 10 mm as maximum settlement. In point I the analysis predicts heave
whereas settlements have been measured, but for point
E settlements are overpredicted.
RESULTS OF 3D MODEL
In this section results from 3D analyses are presented.

These analyses have been performed because the
geometry of the excavation (approximately quadratic
in plan view) and also part of the bracing system (struts
across the corners of the excavation) cannot be adequately represented in plane strain conditions. In the
first series of analyses emphasis has been put on the
stiffness of the diaphragm wall and 3 different calculations have been performed: the first assumed linear
elastic behaviour for the wall with a stiffness assigned
representing stiffness I (uncracked conditions), the
second one assumed stiffness II (cracked conditions)
and the third one introduced a non-linear behaviour by
Figure 7. 2D analysis: surface displacements Point I.
means of a pre-defined curve relating allowable bending moments to the curvature of the wall. In Figures 10
and 11 these are denoted with Z1, Z2 and non-linear
respectively. It has been observed already in the 2D
analyses that the assumption for the stiffness of the jet
grouted panel has as expected a significant influence on the curvature of the wall. The inclinometer
measurements indicate that the lower value obtained
from laboratory experiments seems to underestimate
the support in situ. This has been confirmed also from
196
Figure 8. 2D analysis: surface displacements Point H.
Figure 9. 2D analysis: surface displacements Point E.
Figure 11. 3D analysis: wall deflection right wall.
Figure 10. 3D analysis: wall deflection left wall.
3D analyses and therefore only results assuming the

high stiffness (1,500 MPa, as used in V3 and V4 of the
2D calculations) are presented in the following.
The comparison of horizontal displacements (Figures 10 and 11) clearly show the effect of varying the
stiffness of the diaphragm wall in the unsupported
zone whereas the assumption of cracked stiffness
is closer to the measured curvature than the analysis with high wall stiffness, at least for the right wall.
In the upper part the influence of varying wall stiffness is much less pronounced because the behaviour is
dominated by the struts, however predicted horizontal
displacements are less than measured. The non-linear
model is, not surprisingly, between the two extreme
cases.
Finally, after some discussion with the designer,
an additional analysis was performed assuming a
non-perfect connection of struts and wall, i.e. it was
assumed that there is an imperfection before the full
support of the strut can be mobilised. This has been
achieved by a nonlinear model for the strut which
197
Figure 13. Comparison 2D-3D analysis right wall.
Figure 12. Comparison 2D-3D analysis left wall.
results in a 0.25 mm/m gap to be closed before the

full support is activated (this variation is denoted as
V7 in the following diagrams). The consequence of
this follows form Figures 12 and 13, in which results
from the 2D analysis (Variation 4) are also plotted for
comparison. For the left wall the curvature at the position of the grout panel is still not in full agreement
with measurements but the upper part corresponds
much better than in previous analyses. For the right
wall the curvature and the upper part are now in reasonable good agreement with measurements (for the
right wall the 2D analysis is also in good agreement).
Figures 14 to 17 plot settlements at various observed
points. It is immediately noticed that in contrast to
the 2D model the 3D analysis predicts settlements
also for Point I, although they are still slightly lower as
compared to measured values. Point H corresponds in
the sense that measured and calculated settlements are
almost zero. Point E shows slightly higher settlements
for later stages of construction than measured and the
same holds for point G.
Figure 14. Comparison 2D-3D analysis: Point I.
Figure 15. Comparison 2D-3D analysis: Point H.
198
jet grout panel have been varied. The study showed that
a 2D analysis would reasonably predict wall deflections (in particular for the right wall) but if both walls
and vertical displacements of all surface points are
considered the 3D analysis produces a somewhat better
overall agreement with the measurements.
REFERENCES
Brinkgreve, R.B.J. 2002. PLAXIS, Finite element code for soil
and rock analyses, users manual. Rotterdam: Balkema.
Schweiger, H.F. & Breymann, H. 2005. FE-analysis of
five deep excavations in lacustrine clay and comparison
with in-situ measurements. In (K.J. Bakker, A. Bezuijen,
W. Broere, E.A. Kwast, eds.), Proceedings 5th Int. Symp.
Soft Ground, Amsterdam, June 1517, 2005, Taylor &
Francis/Balkema, Leiden, 887892.
Figure 16. Comparison 2D-3D analysis: Point E.
Figure 17. Comparison 2D-3D analysis: Point G.
CONCLUSIONS
Results form 2D and 3D finite element analyses of

a deep excavation have been compared to in situ
measurements. The excavation is supported by a
diaphragm wall, 3 rows of struts and a jet grout panel
located just below the final excavation depth. In a parametric study the stiffness of the diaphragm wall and the
199
The effect of deep excavation on surrounding ground and nearby structures

A. Sieminska-Lewandowska & M. Mitew-Czajewska
Warsaw University of Technology, Warsaw, Poland
ABSTRACT: In the paper problems related with the execution of 29 m deep excavation of Nowy Swiat Station
(S11) of 2nd metro line in Warsaw are discussed. In the central section, Warsaw 2nd metro line runs below the
center of the city (office and housing buildings and high traffic roads) as well as below Vistula river. This central
section consists of 7 stations and 6 running tunnels 6 km length in total. Running tunnels will be constructed
using TBM, stations cut and cover method. Deep excavation will be executed within diaphragm walls. The
stability of the walls will be provided by several levels of slabs and struts. The analysis of settlements of ground
surface, surrounding foundations and displacements of walls of the excavation have been made. Additionally,
settlements of the surface were calculated above the TBM (running tunnels). Resulting values of settlements in
both cases were compared and discussed.
INTRODUCTION
Construction of 2nd Metro line in Warsaw is scheduled

to begin in January 2008, announcement of design and
build tender has been already published. In the central
section, Warsaw 2nd Metro line crosses below the center of the city (office and housing buildings and high
traffic roads) as well as belowVistula river.This central
section consists of 7 stations and 6 running tunnels
6 km length in total.
Running tunnels will be constructed using TBM,

stations cut and cover method. In the paper problems
related with the execution of 29 m deep excavation in
Quarternary soils (silty sands, sands, clayey sands and
Pliocene clays) are discussed. Within the excavation
Nowy Swiat Station (S11) of the 2nd metro line will
be built. The S11 station will be founded at the depth of
29 m below ground surface (b.g.s.) in the vicinity of so
called Warsaw Slope, where the denivelation (difference in ground surface levels) reaches 30 m (Figure 1).
Figure 1. Longitudinal section of the central part of the 2nd Metro line in Warsaw.
201
Figure 2. Longitudinal section of the central part of the 2nd Metro line in Warsaw.
The depth of the station is a consequence of sudden lowering of the tunnel from the upper slope level
to the level below the bottom of the river with the
consideration of appropriate soil cover resulting from
TBM technology. Deep excavation will be executed
within 100 cm thick diaphragm walls. The stability of
the walls will be provided by several levels of slabs
and struts. In the close vicinity of the excavation and
above the tunnel there are many old buildings, such as:
historic buildings built in XIXth century, partially
destroyed during the 2nd World War and rebuilt
after the war. Polish Academy of Sciences, Warsaw
University and a Hospital are located there. These
buildings are founded on spread foundations at a
depth of 4,80 m b.g.s. Shortest distance between
the excavation wall and foundation of the building
amounts to 3 m;
residential buildings constructed in 30. of XXth
century, probably founded on piles. These buildings
are located above the tunnel drilled using TBM;
residential and office buildings constructed in 50.
and 60. of XXth century on old pre-war foundations. These buildings are founded on spread
foundations at a depth of 4,00 m b.g.s., 6 m apart
from the excavation wall;
masonry and concrete residential and office buildings constructed in 60. of XXth century, founded
at a depth of 6,00 m b.g.s., 5 m apart from the
excavation wall.
Polish Central State Bank and the Ministry of

Finance are located there.
Theses buildings are mostly masonry or reinforced
concrete structures in good technical state. The majority of them is protected by the heritage conservator
law. The location of the excavation of S11 Station,
running tunnels and surrounding buildings is shown
on Figure 2.
The analysis of settlements of ground surface, surrounding foundations and displacements of excavation
walls have been made. Additionally, settlements of the
surface were calculated above the TBM, T11 running
tunnel (cross-section marked by green line). Figure 2
shows the location of all calculation cross-sections in
the vicinity of S11 Station and T11 running tunnel.
GEOLOGY
There are Quaternary and Tertiary soils in the area of

the deep excavation of the station and running tunnels.
According to the geotechnical investigations report,
following geotechnical layers are distinguished:
202
layer I uncontrolled fills 1,52 m thick, in some

places up to 3 m;
layer II moraine deposits reaching depth of 4 m
b.g.s., consisting of medium and stiff sandy clays
and clayey sands of Warta glaciation;
Vertical spacing of slabs is 3 m, which gives an opportunity to adopt underground surface for car parks and
retail. Construction stages are considered as follows:
execution of guide-walls, 1 m thick diaphragm
walls and 1 m high reinforced conrete girt on the
entire perimeter of the excavation,
excavation till the depth of 2 m b.g.s., i.e. below
the slab at level 1, execution of barrettes and
temporary slab supports,
construction of the slab at level 1, backfilling the
excavation and allow traffic back,
excavation till the depth of 5 m b.g.s., i.e. below the
slab at level 2,
construction of the slab at level 2,
excavation till the depth of 8 m b.g.s., i.e. below the
slab at level 3,
the slab at level 4,
excavation till the depth of 26,5 m b.g.s.,
installation of temporary struts at the depth of 26 m
b.g.s.,
final excavation till the depth of 29 m b.g.s.,
construction of 1,5 m thick foundation slab,
deinstallation of the temporary struts.
Figure 3. Calculation section N 3.
layer III medium sands and silty sands of Odra

glaciation, to the depth of 10 m b.g.s.;
layer IV moraine deposits reaching depth of 13 m
b.g.s. consisting of medium and stiff sandy clays of
Odra glaciation;
layer V pliocen clays till the depth of 50 m b.g.s.
There are three levels of ground water table. Considering temporary stability of the bottom of the
excavation, it was assumed that the water table would
be lowered during construction of the station. Geotechnical conditions, distribution of soil layers and location
of foundations are shown on Figure 3 and Figure 4. S11
station and T11 running tunnel are both located within
the layer of stiff and very stiff Pliocene clays.
3
DESCRIPTION OF THE DEEP EXCAVATION

OF THE S11 STATION
Calculations were made in 3 sections, chosen

because of the vicinity of significant buildings.
Figure 3 presents example cross-section N 3,
located close to the beginning of the running tunnel (for the location of the section refer to Figure 2),
showing geotechnical conditions and surcharges.
4
It was designed that the deep excavation of S11 Station will be executed within 100 cm thick diaphragm
walls, founded 10 m below the bottom of the excavation (that means the height of walls is 39 m). Due to the
great depth of the excavation, amounting to 29 m, slab
method of the execution of the excavation was chosen
in order to provide maximum safety of the construction
works. The stability of diaphragm walls will be provided by 8 levels of 35 cm thick underground slabs.
DESCRIPTION OF THE T11 TUNNEL
Two versions of the tunnel structure has been

considered: 1 tube including 2 tracks and 2 tubes, single track each.
The lining of the tunnel was assumed to be constructed of 40 cm thick segments. Following stages of
the execution of the tunnel were modeled:
203
initial stress including overburden and surcharges

(buildings and traffic),
excavation of the tunnel and construction of the

lining of the tunnel.
5.1 Calculations of the excavation of S11 station
its vicinity to the T11 running tunnel. Geotechnical

conditions and location of existing buildings has been
presented on Figure 3, FEM model is shown on Figure 5. Figure 6 presents maximum deformations of the
model in the final construction stage. Maximum calculated lateral displacement of the diaphragm wall in
section 3 amounts to 49,3 mm; maximum foundation
displacement 30,6 mm.
Table 1 presents maximum calculated values of
horizontal and vertical displacements of the wall as
well as settlements of the surrounding buildings in 3
cross-sections chosen for calculation.
Finite element plain strain analysis were carried out

using PLAXIS v. 8 software, Coulomb-Mohr constitutive soil model was chosen for modeling the soil
body, diaphragm walls as well as slabs were modeled as
3-nodes, linear beam elements. Non-associated plastic
flow law was considered. For modeling wall frictions
Coulomb-Mohr low was used. Model dimensions are:
65 m (vertical), 100 m (horizontal), they were estimated taking into account polish regulations according
to the range of influence zone of the excavation.
FEM model mesh, generated automatically, was
built of 807 15-nodes triangle elements and 9773
nodes. For the purpose of the paper 3rd cross-section
was chosen to be presented and discussed because of
Figure 5. Numerical model section 3 (PLAXIS).
Figure 4 shows calculation cross-section N I-I (for

the location of the section refer to Figure 2) including
geotechnical conditions, tunnels (2 tubes) and location
of existing buildings.
CALCULATIONS
Figure 4. Calculation section No I-I (T11).
204
5.2
Calculations of the T11 running tunnel

nodes.
Place Finite element plain strain analysis were carried

out using GEO4 TUNNEL software, Coulomb-Mohr
constitutive soil model was chosen for modeling the
soil body, tunnel lining was modeled using 3-nodes,
linear beam elements. Non-associated plastic flow law
was considered. For modeling wall frictions CoulombMohr low was used.
Section I-I, 1 tube, 2 tracks:
model dimensions: 60 m (vertical) and 240 m
(horizontal);
Section I-I, 2 tubes, single track each:

model dimensions: 60 m (vertical) and 240 m
(horizontal);
nodes.
Geotechnical conditions, tunnels location (2 tubes,
single track each case) and location of existing buildings has been presented on Figure 4, corresponding
FEM model is shown on Figure 7. Figure 8 presents
maximum deformations of that model.
Table 2 presents maximum calculated values of
bending moments, and displacements of the lining
as well as settlements of the surface and surrounding
buildings.
6
Figure 6. Final displacements section 3 (PLAXIS).

Table 1.
Results of calculations of deep excavation S11.

Maximum
displacements of
diaphragm wall
Maximum
settlements
of buildings
Section
Ux
[mm]
Uy
[mm]
U
[mm]
1-1
2-2
3-3
46,1
61,0
49,3
32,8
52,8
53,3
24,5
35,2
30,6
CONCLUSIONS
Taking into consideration results of analysis of the

excavation of S11 Station as well as the results of T11
running tunnel calculations following conclusions are
formed:
1. In the vicinity of the 29 m deep excavation, which
will be executed during the construction of S11
Metro Station estimated settlements of the surface
and surrounding buildings amount to 24,535 mm.
2. Calculated settlements of the ground surface and
surrounding buildings above the T11 running tunnel constructed by the means of TBM, taking into
consideration both cases 1 two track tunnel and
2 single track tunnels are similar and amount to
37,537,8 mm.
3. Theoretical values of settlements as well as displacements and forces in the structures were
Figure 7. FEM model, T11 tunnel 2 tubes, (GEO4 TUNNEL).
205
Figure 8. Maximum deformations of the model, T11 tunnel.
Table 2.
5. During the construction, the results of analysis

described in the paper will be carefully verified and
discussed.
Results of calculations of running tunnel T11.

Finite elements method
Type of
tunnel
Maximum
bending moments,
displacements
of tunnel lining
Maximum Maximum
surface
settlements
settlements of buildings
Mmax
[kNm/m]
Umax
[mm]
Umax
[mm]
Umax
[mm]
24,2
37,7
37,8
8,1
37,5
37,6
1-1
306,6
(1 tunnel)
1-1
290,2
(2 tunnels)
calculated considering that the value of the modulus of deformation of Pliocene clays, within which
the structures are located, is E = 50 MPa. This value
must be verified by means of in-situ tests and then
the calculations will be adjusted.
4. Due to the expected differences in the values of
settlements of the ground surface close to the deep
excavation and above the tunnel further analysis
of the case including 3D modeling of the contact
of 2 types of tunnel structure (running tunnel and
station) will be performed.
REFERENCES
FINE Ltd. 2007. GEO4 Users manual. Prague: FINE Ltd.
Geoteko Sp. z o.o. 2004. Evaluation of the technical state
of buildings in the influence zone of 2nd metro line in
Warsaw, section Rondo Daszynskiego Station Powisle
Station. Warsaw: Geoteko Sp. z o.o.
Geoteko Sp. z o.o. 2004. Geotechnical and hydrological
report for the construction of 2nd metro line in Warsaw,
Nowy Swiat
Station. Warsaw: Geoteko Sp. z o.o., SGGW,
Geoprojekt Sp. z o.o.
Grodecki, W., Sieminska-Lewandowska, A. & Lejk, J. 2007.
Second metro line in Warsaw possibility and methods of realization, Inzynieria i Budownictwo 7-8/2007:
365368.
Kotlicki, W. & Wysokinski L. 2002. Protection of structures
in the vicinity of deep excavations (376/2002). Warsaw:
Building Research Institute.
PLAXIS BV. 2005. PLAXIS Users manual. Roterdam: A. A.
Balkema.
Polish Committee of Standardisation. 2002. PN-EN 15382002 Execution of special geotechnical works. Diaphragm
walls. Warsaw: Polish Committee of Standardisation.
206
Multi-criteria procedure for the back-analysis of multi-supported

retaining walls
J. Zghondi
Arcadis, Lyon, France and LGCIE, INSA-Lyon, France

LGCIE, INSA-Lyon, France
ABSTRACT: A numerical back-analysis procedure for multi-supported deep excavations is proposed based
on the optimization of several indicators, taking in account the forces in the struts and the differential pressures
derived from the wall displacement. The evaluation of the procedure is performed on 1 g small scale laboratory
experiments (Masrouri 1986) on semi-flexible retaining walls embedded in Schneebelli material. The proposed
numerical procedure was applied on an excavation with 2 passives low stiffness struts. The resulting Hardening
Soil Model parameters are further used to back-calculate the 14 different tested configurations. The results are
compared with the classical methods, SubGrade Reaction Method, Finite Element analysis with Mohr Coulomb
model with parameters proposed by Masrouri (1986) and with the back-analysis using Hardening Soil Model
parameters based on triaxial tests results.
INTRODUCTION
Numerical back-analysis of in situ monitoring results

of multi-supported deep excavations is generally
extremely complex (Hashash & Whittle 1996, Finno &
Calvello 2005, Delattre 1999): soil characteristics can
be heterogeneous or determined with a low degree of
confidence, the different stages of the excavation can
be difficult to reproduce in a 2D numerical approach,
key mechanical parameters can be unknown (for example the actual stiffness of the strut-to-wall contact)
and the number of measured quantities such as wall
displacements, settlements and strut forces is generally too small to perform a comprehensive comparison
between the actual behavior and the numerical results.
Thus the full validation of a back-analysis numerical procedure (including in particular the choice of
the constitutive law and the determination of all the
required parameters) is rarely directly possible on real
case histories. Therefore, the numerical procedure proposed in this paper is validated on 1 g small scale laboratory experiments performed by Masrouri (1986) on
semi-flexible retaining walls embedded in Schneebeli
material (mixture of steel rods of different diameters
representing in 2D the behavior of a cohesionless soil).
The 14 considered experiments correspond to a retaining wall, whose length and mechanical properties are
kept constant, supported by one or two levels of active
or passive steel struts with various axial stiffness and
prestressing.
Even for such simple comprehensive laboratory

experiments, usual design method like SubGrade
Reaction Method (SGRM) or classical limit equilibrium methods do not capture all the observed behaviors
and test results. It is thus necessary to propose a unified numerical procedure to back calculate with an
acceptable degree of confidence, all the results of the
excavations tests of Masrouri (1986).
2
EXPERIMENTS AND COMPARISON WITH

CLASSICAL/SGRM CALCULATIONS
The experiments correspond to small scale 2D models

of flexible retaining walls (Figure 1).
Figure 1. Masrouris experimental set up.
207
Table 1.
Summary of Masrouri (1986) experiments.

First strut
Second strut
Experiments
Prestressed
(kN/m.ml)
Stiffness
(kN/m.ml)
Prestressed
(kN/m.ml)
Stiffness
(kN/m.ml)
B1
B2
B3
B4
B5
B6
B7
B8
B10
B11
B12
B13
B14
2.133
3.025
0.208
2.1
3
0.191
2.016
2.916
0.383
1.330
0.366
2.075
1.733
83333
83333
816
816
816
404
404
404
83333
83333
816
816
816
(not used)
(not used)
(not used)
(not used)
(not used)
(not used)
(not used)
(not used)
0.333
2.691
0.400
2.700
4.041
83333
83333
816
816
816
Schneebelli 2D analogic soil was used: this material offer a good repeatability and enables to build
a homogeneous 2D soil model with quick handling
for the experiments. However Schneebeli materials
have some inconvenients: the unit weight is close to
6.5 kN/m3 , the angle of friction is smaller than that of
most of the soils (21 ) and it only presents a dilatant
behavior.
While maintaining the same geometrical and
mechanical characteristics for the wall (EA = 1.2
106 kN/m and EI = 14.4 kN/m2 /m), a wide range of
configurations (1 or 2 struts, with different prestressing and stiffness) was considered, see Table 1. The
phases used in all the excavations were planned as
follows;
1. strut cases: 10 cm of excavation, installation of the
first strut at 5 cm from the top and prestressing, then 3 excavations of 10 cm each till 40 cm,
then several excavations of 5 cm until failure is
obtained.
2. struts cases: the same procedure as for the 1
strut cases is followed until the excavation reaches
40 cm from the top, then the 2nd strut is installed
at the level 25 cm and prestressed. The step-wise
excavation (by increment of 5 cm) is resumed until
failure occurs.
The strut to wall contact is hinged in order to prevent
bending moment to be transmitted to the struts. For
each excavation phase, the horizontal displacements of
the top or bottom of the wall are measured. The curvature is also measured in 26 locations (both sides on the
wall) allowing to determine with a reasonable accuracy
the differential pressure acting on the wall from the top
to the bottom (polynomial approximation performed
by the Palpan program created by Boissier et al. 1978).
Photographs were also taken to determine the displacement fields with a stereophotogrammetric technique.
80
70
60
B3
Passive strut
low stiffness
50
exc -45cm
40
30
20
10
-15
Experiment
SGRM
0
-5
5
15
Differential pressure (kN/m2)
Figure 2. Differential pressure calculated with SGRM compared to experimental values: Case B3 low stiffness
passive strut (45 cm excavation level).
Load cells determine the forces on the cylindrical

struts.
2.1 Classical and SGRM calculations
The whole series of experiments were first compared
with classical (modified Blum method) and SGRM
calculations (Terzaghi 1955). The subgrade reaction
modulus used in RIDO calculations (Fages 1996) is
increasing with depth in the following manner:
K h = K + K v .
with K = 0,
Kh = K v
Classical methods calculation totally neglects the
influence of the stiffness of the wall, construction
steps, stiffness and prestressing on the struts, arching
effect, etc. On the opposite SGRM explicitly considers
the stiffness of the wall, the construction steps and the
stiffness of the strut.
In the case of one passive strut with low stiffness
(tests B3, B6), the SGRM method reproduces in an
acceptable manner the differential pressures results
(Figure 2).
208
80
70
60
80
B10
Passive stiff
strut
70
60
50
50
40
exc -45cm
exc -45cm
30
40
30
20
20
Experiment
10
SGRM
0
-20
B14
Active low
stiffness
-10
0
10
20
Differential pressure (kN/m2)
Experiment
10
SGRM
0
-20
-10
0
10
20
Differential Pressure (kN/m2)
Figure 3. Calculated (SGRM Method) and measured differential pressure for B10 and B14 at 45 cm of excavation.
Figure 5. Strut forces for cases B1 to B8 with 1 strut

variation during the last 4 excavation phases before failure.
Deviatoric Stress kPa
600
500
Experiments
HSM1
400kPa
kPa
400
400
300kP
kPa
300
a
300
200kPa
kPa
200
200
100
0
0,00
1,00
2,00
Figure 4. Mesh used for the Plaxis calculations of 2 struts

supported walls.
3,00
4,00
5,00
Axial Strain %
12
Volumetric strain/1000
With one stiff or active strut or in the case of excavations with 2 struts, both the SGRM and the limit
equilibrium methods fail to reproduce the experiment
differential pressures (Figure 3a Test B10). In some
cases, a good description of the pressure diagrams
seems to be obtained (Figure 3b Test B14). It actually
results from two errors compensating each other: the
over estimation of the pressure induced by prestressing
and the lack of ability to reproduce the arching effect.
Experiments
HSM1
10
8
500Kpa
200Kpa
6
4
200Kpa
400Kpa
0
-2
-4
Axial Strain%
2.2
Finite element back calculation with Mohr

Coulomb model
Figure 6. Simulation with HSM1 parameters of the biaxial

tests with 200, 300 and 400 kPa.
It appears that the classical and SGRM methods do not,

even in simple cases of excavation, accurately describe
all the observed results. A finite element approach is
therefore proposed.
The finite element calculations were performed
with Plaxis V8.2, the model represents a vertical slice
of Masrouris experiment. The mesh is composed of
triangular 15 nodes elements (Figure 4).
For these calculations, the same characteristics of
the Schneebelli material were used (c = 0, = 21 ,
= 65 kN/m3 ), taking into account for the soil-wall
interface an interface factor Rint = 0.55 (related to the
friction angle value noted by Masrouri).
Masrouri (1986) estimated the elastic modulus at
4500 kPa with an increment of 26830 kPa/m after
0.6 m of depth (values back calculated on test B4 with

1 active strut of low stiffness).
In all the presented calculations the difference with
the measured strut forces and wall displacements do
not exceed 20% (see Figure 5 for the 1-strut cases B1
to B8).
A sensibility analysis is performed considering
all the parameters of the Mohr Coulomb model. It
appears that the differences between calculated and
experimental results can not be satisfyingly reduced
(especially in the cases with 2 struts). Therefore a more
sophisticated constitutive law is required (Figure 6):
the Hardening Soil model.
209
Figure 8. Differential pressure curves for indicator

explication.
Figure 7. Differential pressure for test B12 and excavation
at 50 cm: comparison of the experimental values with the
results of SGRM, FE with Mohr Coulomb law and Hardening
Soil Model (HSM1 and HSM2).
2.3
FE back-calculation with Hardening soil model

parameters based on triaxial tests
Hardening soil model (Shanz et al. 1999) can capture

soil behavior in a very tractable manner. The values of
the different parameters were first fitted on the biaxial
tests results (under a 200 Kpa confining pressure) performed on 20 10 cm samples by Kastner (1982). The
200 kPa confining stress is greater than the mean stress
generally observed in Masrouri experiments, but in a
first approach it was considered as more representative
of the soil behavior than the biaxial tests using lower
confining pressures: the size of the sample and the
high value of the unit weight of Schneebelli material
induces a non homogeneous stress state in the sample
that could greatly affect the accuracy of the results.
The obtained set of parameters noted HSM1 in the
sequel provides a satisfying description of the biaxial
test with 200 kPa or more confining stress (Figure 6).
This set of parameters was further used to back calculate the whole series of Masrouris experiments (14
cases). It appears that these calculations do not reproduce well the test results in term of strut forces or
differential pressures on the wall (Figure 7 represents
only the results of differential pressures).
3
PROPOSED BACK ANALYSIS PROCEDURE
The aim of this procedure is to find the proper set

of parameters for the constitutive soil model (Mohr
Coulomb or Hardening Soil models) considered in the
Finite Element simulations. Considering one particular test configuration, the parameters will be first
obtained from the minimization of indicators based
on differential pressures and struts forces errors.
The resulting set of parameters will then be confronted with the results of the biaxial tests and of 14
configurations considered experimentally by Masrouri

(1986).
3.1 Definition of indicators
In order to consider the main features of the retaining
wall behaviour, two indicators were defined: Esm is
related to the strut forces and Epd related to the differential pressure on the wall (linked to the displacement
profile).
where Eb1 = f1 f1 and Eb2 = f2 f2 and with f1(2) and

f1(2)
the strut force calculated or measured in the 1st
(2nd) strut level.
The Esm indicator is based on the error of the sum
of struts forces: therefore an error on the strut force f1
can be compensated by f2 .
The Epd indicator (Figure 8) takes into account the
absolute value of the difference between the measured
and calculated differential pressures (respectively P1
and P2 ), divided by the integral of the measured differential pressure P1 . Integrals are calculated from the
top of the wall to 10 cm below the final excavation
level.
Epd is the main indicator while the possible inaccuracy of the strut force measurement especially at the
beginning of the excavation makes the Esm indicator a
second validation indicator.
3.2 Parameter optimization for Mohr Coulomb
model based on a 2 strut excavation test
The determination of a second set of parameters for
Mohr Coulomb model is performed by fitting the
final results of a 2 struts excavation case (excavation
level 50 cm). The particular case of test B12 (corresponding to passive struts with low stiffness) was
selected because it clearly appears that the SGRM fail
210
Figure 9. Variations of indicators Esm and Epd with Eref

50 .
to reproduce the differential pressure diagram obtained

in such configuration.
Only one independent parameter was used for fitting the results, the elastic modulus E. The other
parameters are kept constant (c, , . . .). The final
value of E is determined through the optimization of
the indicators Esm and Epd . Actually, a linear variation
of E with the depth is considered:
Both the initial value of the elasticity E0 and its

variation with depth were considered in the optimization procedure. It appeared that none of these
parameters could be modified to improve the description of the experimental results, indicating that the
Mohr Coulomb model is unable to reproduce important features of the soil behaviour involved in the global
behaviour of the retaining wall.
3.3
HSM model optimized by fitting on a 2 strut

excavation test
The determination of a second set of parameters for

Hardening Soil Model (noted HSM2 in the sequel) is
performed by fitting the final results of the same 2
struts excavation case (B12 excavation level 50 cm)
as in section 3.2.
Only one parameter was used for fitting the results,
ref
the reference stiffness modulus E50 . The other parameters are either constant (c, , ) or keep a direct
ref
relationship with E50 . For example:
ref
The final value of E50 is determined through the optimization of the indicators Esm and Epd . Considering
the accuracy of the experimental results (in particular the procedure leading to the differential pressure
ref
diagrams), E50 is determined with a precision of
1000 kPa (Figure 9).
ref
The value of E50 = 6000 kPa is chosen because it
appears to minimize both indicators. HSM2 is further
used to simulate the biaxial tests with low confining
stress (50 and 100 kPa). Figure 10 shows a good agreement with the experimental results, even though the
latter can be affected by the non-homogeneity of the
initial stress state in 20 cm 10 cm samples.
Figure 10. q-1 biaxial curves: experimental results and

calculated values with HSM2.
The proposed back analysis procedure shows its

ability to verify or justify the HSM parameters that
accurately describe the biaxial test results.
3.4
Back calculation of the 14 configurations tested

by Masrouri (1986)
The HSM2 set of parameters is now used to back calculate the 14 different tests (Table 1). Figures 11 and
12 present the differential pressure diagrams obtained
with HSM2 and other methods and the ratios f /f of
the calculated to the measured strut forces respectively
for tests B1 to B8 (1 strut) and tests B10 to B13 (2
struts).
The strut forces and the differential pressure are
well represented compared to the classical methods,
the SGRM method or Finite element analysis with
a simpler constitutive model (Mohr Coulomb). The
effect of the prestressing of the strut is well reproduced
by that procedure (comparing B3 and B5 in Figure 11
or B10 and B11 in Figure 12), as well as the arching
effect (case of B10 in Figure 12) and the influence of
the strut stiffness (comparing B1 and B7 in Figure 11).
The proposed back analysis procedure and the
resulting set of parameters HSM2 show their efficiency in all the configurations (unlike the SGRM
where the arching effect and the prestressing on the
struts are not well reproduced).
3.5 Summary
Figure 13 presents a comparison of the values obtained
for the selective indicator Epd in 3 cases of excavation
and for the SGRM, Mohr Coulomb, HSM1 and HSM2
models:
B3 corresponds to a single passive and low stiffness
strut
B12 uses 2 passive low stiffness struts
and B10 2 passive rigid struts.
In all of these 3 cases, it appears that the HSM2
set of parameters clearly minimizes the error between
experimental and numerical results.
211
Figure 12. Differential pressure diagrams and strut forces

obtained with HSM2 (excavations with 2 struts at 50 cm)
compared with experiments and classical methods results.
Figure 13. Epd indicator calculated for tests B3, B12, B10
(excavation 50 cm), for the SGRM, HSM1, HSM2 and
Mohr Coulomb calculation.
With the same constitutive model (HSM1 and

ref
HSM2) and with the same parameters except E50 , the
error is divided by 3 to 5.
4
Figure 11. Differential pressure diagrams and strut forces
obtained with HSM2 (excavations with 1 strut at 50 cm)
compared with experiments, Mohr Coulomb, and classical
methods results.
CONCLUSION
A comprehensive series of 14 small scale experiments on flexible retaining walls with different strut
212
stiffness and prestressing is used to validate a numerical back calculation using the Hardening Soil Model.
The final set of parameters HSM2 is fitted on one
single test (B12 at the final excavation level). The
proposed model is based on the simultaneous minimization of two indicators Esm and Epd respectively
related to the strut forces and differential pressure diagram. The verification of the proposed back analysis
procedure showed that the HSM2 model gives the most
acceptable description of the differential pressures and
forces on the struts in all of the 14 tested configurations
compared to the SGRM method or a Finite Element
approach with either Mohr Coulomb model or Hardening Soil model with parameters based on biaxial tests
(HSM1).
Further developments will include the verification
of the ability of proposed back-analysis procedure to
determined the HSM parameters not only on Esm and
Epd at the final excavation level but also on intermediate levels, for example the first excavation step after
the installation and prestressing of the lower strut.
Despite the already mentioned difficulties, further validation of the procedure on well-instrumented
excavation sites will also be tested.
Brinkgreve, R.B.J. & Broere, W. 2004. Plaxis V8 manuel.

Delattre, L. 1999. Comportement des crans de soutnementExprimentations et calculs. PhD dissertation, ENPC,
Paris, ENPC, 498 p.
Fages, J. 1996 Rido Users manual, RFL, Miribel, France.
Finno, R. & Calvello, M. 2005. Supported excavations:
observational method and inverse modeling. J. Geotech.
Geoenv. Eng. ASCE, 131, (7): 826836.
Hashash, Y. M. A. & Whittle, A. J. 1996. Ground Movement
Prediction for Deep Excavations in soft Clay. Journal of
Geotechnical Engineering, 122(6): 474486.
Kastner, R. 1982. Excavation profonde en site urbain Problmes lis la mise hors deau- Dimensionnement des
soutnements butonns. PhD dissertation, INSA Lyon &
Universit Claude Bernard-Lyon 1, 409 p.
Masrouri, F. 1986. Comportement Des Rideaux de soutnement Semi- Flexibles. PhD dissertation, INSA Lyon,
France.
Schanz, T., Vermeer, P.A., Bonnier, P.G., 1999. The
Hardening-Soil Model: Formulation and verification.
In: R.B.J. Brinkgreve, Beyond 2000 in Computational
Geotechnics. Balkema, Rotterdam: 281290.
Terzaghi, K. 1955. Evaluation of coefficients of subgrade
reaction, Gotechnique, 5: 297326.
REFERENCES
Boissier, et al. 1978. Dtermination des moments et des
pressions exercs sur un cran partir de mesures inclinomtrrique. Revue Canadienne de Gotechnique, 15, (4),
522536.
213
Monitoring and modelling of riverside large deep

excavation-induced ground movements in clays
D.M. Zhang & H.W. Huang
Key Laboratory of Geotechnical and Underground Engineering of Ministry of Education, Tong ji University,
Department of Geotechnical Engineering, Tong ji University, Shanghai, P.R. China
W.Y. Bao
China State Construction Engineering Corporation (SH), Shanghai, P.R. China
ABSTRACT: The Riverside large deep excavation of Shanghai international passenger center was 800 m long
and 100150 m wide with the depth of 13 m. The south long side of the deep excavation was at a distance of
4.6 m from the parallel flood wall of Huangpu River. The north long side was 5 m away from a historic building.
Problems resulted from the large deep excavation was the asymmetric ground movements along the long sides
due to the complex surrounding condition and surface surcharge. The monitoring during the excavation provided
numerous data to study the characteristics of the ground movement and earth pressure. The numerical modelling
was also adopted aim to predict the ground movements.
INTRODUCTION
The development of underground space along the bund

of Huangpu River in Shanghai, China has resulted in
excavations becoming progressively larger and closer
to the River, where the groundwater table was just near
the ground surface and a great number of underground
works are within a few meters of the surface. The
riverside excavations were all located close to the existing buildings, network and the city lifeline of flood
wall. It has become a great challenge to protect these
neighboring buildings and public utilities from damage during the deep excavation due to the complex
geotechnical constraints and the small opening from
the Huangpu River. The soils near the Huangpu River
was usually weak with a very low strength and higher
water content, which were a potential causes of the
larger ground movement. Meanwhile, the complex and
dense environments put forward a strict requirement
on the ground movement controlling. It was difficult
to determine the earth pressure acted on the retaining
wall with any conventional earth pressure theory considering the small soil body left between retaining wall
and flood wall. Besides, the retaining wall of riverside
deep excavation was usually asymmetrically loaded
with much higher earth pressure on one side, which
was caused by great surface surcharge due to the existing buildings and the pile of the construction material.
The stability of the deep excavation as a whole was
worth considering to avoid any kinds of failure of

the deep excavation and consequent damage on the
environments.
However, there were few references for the construction of the large deep excavation because of
the geotechnical condition and complex environment
along the bund of Huangpu River. The deep excavation of Shanghai international passenger center (SIPC)
was the largest and closest one to Huangpu River so far.
The construction and the analysis method of the deep
excavation of SIPC and the induced ground movement
as well will be a useful and practical reference for the
subsequent riverside large deep excavation.
2
2.1
PROJECT OUTLINE AND SOIL CONDITIONS

Project outline
The deep excavation of SIPC was 800 m long with the

width of 100150 m and the depth of 13 m. The large
deep excavation was divided into two sub-excavations
with the lengths of 480 m and 218 m respectively to
reduce the risk of damage for the existing structure and
the failure of deep excavation. The study presented in
this paper was carried out based on the deep excavation
with the length of 480 m. The sketch view of the project
was illustrated in Figure 1. The space between the deep
excavation and flood wall of Huangpu River was only
215
Figure 1. Sketch view of the project.
Figure 3a. A-A cross section of deep excavation.
Figure 2. The plane view of strut arrangement.
4.6 m at south side. A number of existing buildings,

including a historic building, were located along the
north side of the deep excavation with a distance of
about 5 m.
Bored piles supplemented by SMW piles wall were
used as the retaining structure. The bored pile was
950 mm in diameter with a center-to-center space of
1150 mm. The effective length of the bored pile was
26 m and the embedment was adequate to provide sufficient passive earth pressure to keep the stability of
the retaining wall. The SMW piles were 850 mm in
diameter with an effective length of 20.8 m. The distance between SWM piles was 600 mm to guarantee
the waterproof performance. The mix ratio of cement
was as high as 20% for SMW piles. Three reinforced
concrete struts were set at the depth of 0.9 m, 5.7 m
and 9.6 m with the cross section of 1250 800 mm
for the first strut and of 1200 800 mm for the second
and third strut. The plane space of the strut was about
1.2 m and illustrated in Figure 2. The cross sections
of the deep excavation were presented in Figure 3a
and 3b.
The jet grouting belt of 4 m wide and about 4 m
high was employed closely above the bottom of the
deep excavation along the retaining wall. The grouting
could significantly increase the capacity of the soil
resistance for the retaining wall during the excavation.
The bored pile was extended from 26 m to 27 m near
Figure 3b. B-B cross section of deep excavation.
the historic building to protect the building from crack

and tilting. Besides, the isolation piles were specially
designed to reduce to deep excavation-induced effect
on the historic building.
2.2 Soil conditions
The soil profile throughout the deep excavation comprises the mixed filling to a depth of 6.4 m, which
contains many obstacles and made a lot of trouble for
the deep excavation, underlain by silt, silty clay and
mucky clay. The retaining wall including the waterproof wall was embedded in the silty clay. The detailed
characteristics of the soils were presented in Table 1.
216
Table 1.
Soils characteristics throughout the deep excavation.

Water
content
%
Bulk
density
kN/m3
Compression
modulus
kPa
Cohesion
kPa
Soil
Depth
m
Mixed filling
Silt
Mucky clay with silt
Mucky clay
Silty clay 1
Silty clay 2
Silty clay 3
6.38
4.89
4.96
7.17
7.31
4.05
17.23
31.7
40.5
49.7
34.0
33.3
33.2
18.2
17.5
16.6
18.0
17.8
17.9
8170
4130
2580
4570
8130
5240
8
11
14
17
8
17
Friction
angle
28.5
22
13
17
29
23.5
lateral displacement/mm
0
10
20
30
40
50
depth/m
10
Figure 4. Monitoring layout of lateral displacements.
15
20
Table 2.
11/ 11/ 2005

12/ 2/ 2005
1/ 3/ 2005
1/ 13/ 2006
Progress of excavation.
25
Date
11/5/200511/11/2005
11/11/200512/2/2005
12/2/20051/3/2006
1/13/2006
corresponding
excavation
depth (m)
30
Figure 5. Lateral displacement at inclinometer CX1.
3
6.2
13
completion of
bottom plate
10
20
30
40
MONITORING OF EXCAVATION
3.1
Monitored lateral displacements
Figures 58 showed the lateral displacements at

monitoring points of CX1, CX3, CX12 and CX10
corresponding to the studied excavation stages. These
inclinometers were close to the center of the long side
depth/m
10
As shown in Figure 4, 14 inclinometers, which were

denoted by CX1 to CX14, were set into the retaining wall around the deep excavation. 2 earth pressure
gauges denoted as TY1 and TY2 in Figure 4 were also
installed close to the retaining wall to study the evolution of earth pressure of the small soil body between
retaining wall and flood wall during excavation. The
earth pressure gauges were installed every 5 m in the
vertical overall 25 m.
The progress of excavation was presented in Table 2.
15
20
25
12/2/2005
1/3/2005
1/13/2006
30
of the deep excavation and thus the readings were

representative of the maximum displacement of the
retaining wall. It could be found from figures 58
that the maximum lateral displacement was less than
60 mm during the whole excavation stage. Meanwhile,
217
10 20 30 40 50 60
depth/m
10
15
20
25
11/11/2005
12/2/2005
1/3/2005
1/13/2006
Figure 9. Distribution of lateral movement at the top of

retaining wall.
30
earth pressure/kPa
11/11/2005
5
10
depth/m
depth/m
10
15
25
12/2/2005
1/3/2005
20
50 100 150 200 250 300 350 400
0
10 20 30 40 50
1/13/2006
active pressure
15
20
11/11/2005
12/2/2005
1/3/2005
1/13/2006
25
30
30
Figure 10a. Comparison of recorded earth pressure with

calculated active earth pressure at TY1.
the lateral displacement exhibited asymmetric behavior along the two long sides of the deep excavation
because of the following two reasons. Firstly, the earth
pressure acted on the retaining piles was asymmetric because of the small soil body between Huangpu
River and deep excavation. Secondly, the surface surcharge was asymmetric due to the existing buildings.
Comparing the records of CX1 with CX12, CX3 with
CX10, it could be found that lateral displacement of
retaining wall was 1529% smaller at the south side
than north side. Unfortunately, the larger lateral displacement at north side would result in a potential
damage to the neighboring historic building. Consequently, the jet grouting should be immediately carried
out to improve the foundation of the historic building
and it was proved to be an effective way to avoid the
damage of crack and tilt of the building.
The recorded lateral movements at the top of retaining wall were presented in Figure 9.The five movement
curves from outside to inside were corresponding to

the excavation depth of 2.3 m, 6.2 m, 13 m, completion
of bottom plate and completion of underground structure respectively. The maximum lateral movement at
the top of retaining wall reached 106.5 mm at the north
side, while it was only 50 mm at the south side, when
the underground structure was completed. These phenomena also confirmed the influence of asymmetric
earth pressure on the movement of the retaining wall.
3.2 Evolution of earth pressure
The monitored earth pressure was illustrated in
Figure 10a and 10b with calculated one. The calculated active earth pressure was obtained using Rankine
earth pressure theory.
From figure 10a and figure 10b, it could be found
that the monitored earth pressure at top 15 m was very
218
earth pressure/kPa
0
50 100 150 200 250 300 350 400
0
11/11/2005
5
12/2/2005
1/3/2005
depth/m
10
1/13/2006
active pressure
Figure 11. The FEM mesh.
15
Table 3.
ment.
20
Parameters of retaining piles and ground improve-
25
Retaining piles
Strut
Ground improvement
30
Elastic modulus
kPa
Poisons ratio
3.3 107
3.0 107
1.4 105
0.15
0.15
0.20
Figure 10b. Comparison of recorded earth pressure with

calculated active earth pressure at TY1.
2-D numerical modelling was carried out using FEM

code of Plaxis v8 considering the narrow plane characteristic of the deep excavation. The cross section
11 shown in Figure 2 was adopted in FEM analysis
because it was almost the center of the deep excavation
and near the historic building as well.
large enough to lower the boundary effect. A linear

elastic model was adopted for the ground improvement. The retaining wall as well as the strut was
simplified as elastic beam in numerical modelling.
The soils were simulated with Mohr-coulomb model.
The numerical simulation was performed with 15-node
isoparametric finite elements under the assumption of
plane strain conditions. The FEM model was presented
in Figure 11.
The boundary conditions in the numerical simulation contain the following two types, one was the
displacement boundary condition, and the other was
the drainage condition. A free displacement boundary
condition was adopted at the ground surface. It was
assumed that no horizontal nor vertical displacement
taken place at the lower boundary, for it was beyond
the influence of deep excavation. The lateral displacements at left and right hand boundary were both
fixed as zero. The drainage condition at the ground
surface was assumed to be free, hence the excess pore
pressure was kept as zero along the ground surface;
meanwhile the lower boundary as well as the left
and right hand boundary condition were considered
to be kept as hydrostatic pore pressure during excavation. The initial effective stresses and hydrostatic pore
pressure were calculated based on the weight of the
soil and the underground water condition.
4.1
4.2 Parameters used in numerical modelling
close to the active pressure for both monitored points.

It became much larger than active pressure below 15 m.
Two factors could be contributed to the distribution
of the earth pressure. One was that the magnitude of
lateral displacement of the retaining wall was larger in
the top 15 m than that of below 15 m, and it could be
verified from Figures 58. The other cause was that the
small bulk of the soil body against the retaining wall at
top 15 m. It could be found from Figure 3 that the soil
body in the top 15 m was much smaller than in below
15 m. Both figure 10a and figure 10b implied that the
soil body had a significant effect on the distribution of
earth pressure against the retaining wall.
However, the earth pressure at north side of the deep
excavation was not monitored. No comparison could
be performed between the two sides.
4
MODELLING OF DEEP EXCAVATION
Numerical model
The overall width of deep exaction at cross section 11

was 100 m with excavation depth of 13 m. The width of
the numerical model was 240 m, which was 18 times
as wide as the depth of the excavation. The vertical
dimension was 50 m, which was more than 3.5 times
the depth of the excavation. The model dimension was
The parameters of retaining structure and ground

improvement used in the numerical analysis were
listed in Table 3. The soil parameters could be referenced as Table 1. The interface between retaining piles
and soil was adopted and the interface parameters were
determined according to Plaxis manual. The modulus
219
4.3 Numerical modelling procedure
lateral displacement/ mm
0
10
20
30
40
50
6.2 m
The excavation was modeled with 7 consecutive steps

shown as following: STEP 1 was to determine the initial stress state due to the gravity of soils. STEP 2 was
used to exert the loading of existing buildings on the
surface with the magnitude of 6070 kPa according
to the type of the buildings. The movements induced
in STEP 1 and 2 were reset to zero in the modelling.
STEP 3 represented the construction of retaining wall
and ground improvement, the surcharge of 20 kPa was
also loaded at this step. STEP 4 meant the first excavation to 1.4 m deep and the construction of first strut.
In STEP 5, the second strut was set after excavating to
6.2 m deep. Excavating to 10 m deep and the third strut
was finished in STEP 6. The excavation was completed
and bottom plate was constructed in STEP 7. Dewatering was considered during the excavation by changing
water table.
13 m
4.4 Calculated lateral displacements
depth /m
10
15
20
1.4 m
25
30
Figure 12. Evolution of lateral displacements at south side
of deep excavation.
Figure 13. Evolution of lateral displacements at north side

of deep excavation.
of resilience was adopted for soils. The modulus of

resilience was obtained by back-analyzing the monitored lateral displacements of the first excavation
progress shown in Table 2. It was found that the modulus of resilience was as 5 times high as the compression
modulus for soils.
Figures 1213 presented the evolution of lateral

displacements with excavation progress. The lateral
displacements of south side of the retaining piles
reached 28.8 mm, 38.4 mm respectively when excavated to the depth of 6.2 m and 13 m.They were smaller
than those of north side of the retaining wall, which
were 32 mm and 60 mm. It could be found that the
maximum lateral displacement at north side was 1.5
times larger than that of south side of the retaining wall
by comparing figure 12 and figure 13.
The comparison between calculated and monitored
lateral displacement could be carried out because cross
section 11 was coordinate with inclinometer CX3
and CX10. The monitored lateral displacements were
57.8 mm and 35 mm at monitored points CX3 and
CX10 corresponding to the excavation depth of 13 m,
while the accordingly calculated ones were 60 mm and
38.4 mm respectively. The agreement between calculation and monitoring implied the validation of the
simulation procedure of FEM modelling with backanalysis on the modulus of resilience of soils. Besides,
the consideration of main influential facts, such as surface surcharge due to existing loading and piling of
construction material, the process of excavation, the
supplemented techniques of dewatering and ground
improvement, was essential in the FEM modelling to
reasonably predict the behavior of the retaining piles.
5
CONCLUSIONS
The lateral displacement of Riverside deep excavation

with complex surrounding environment was studied
with monitoring data and FEM modelling. The lateral
displacements of the retaining wall were asymmetric
because of the asymmetric earth pressure. The maximum lateral displacement at north side was almost 1.5
220
times as large as that of south side of the deep excavation. The earth pressure was close to the active earth
pressure in top 15 m due to the large lateral displacement and small soil body against the retaining wall.
The earth pressure was much larger than active pressure below 15 m. It was found the soil body bulk had a
noticeable effect on the distribution of earth pressure
against retaining wall.
REFERENCES
Tunnelling and Underground Space Technology 16:

151158.
Yamagushi, I. & Yamazaki, I. et al. 1998. Study of ground
tunnel interactions of four shield tunnels driven in close
proximity, in relation to design and construction of parallel shield tunnels. Tunnelling and Underground Space
Technology 13(3): 289304.
Zhang, D. M. & Huang, H. W. 2007. Ground movements
and controlling measurements in deep excavation under
asymmetric loading. Proceeding of 10th national conference on soil mechanics. Chongqing, 14 Novermber 2007
(in press). (in Chinese).
Chang, C. T. & Sun, C. W. et al. 2001. Response of a Taipei

rapid transit system TRTS tunnel to adjacent excavation.
221
GPS height application and gross error detection in foundation pit

monitoring
H. Zhang
School of Safety and Resource Engineering, China University of Mining & Technology, Beijing, P.R. China
S.F. Xu
College of Architecture & Civil Engineering, Zhejiang University of Technology, Hangzhou, P.R. China
T.D. Lu
Department of Survey, East China University of Technology, Fuzhou, P.R. China
ABSTRACT: The author introduces a deformation monitoring model combined by traditional measuring technology and modern GPS measuring technology based on technical attribute of foundation pit deformation
monitoring and authors experience of deep foundation pit construction project of underground tunnel in Lishui
Road, Hangzhou city. When analyzing GPS height conversion, in order to improve GPS datum marks reliability,
one can use Dixons test in GPS datum mark reliability test to find out height anomaly, thus provide convenience to search and delete marks with gross error. This test also improves deformation monitoring processs
efficiency.
BACKGROUND PROJECT INTRODUCTION
Lishui Road (from Huzhou Road to Qingfang Road)

project is one of Hangzhou City governments 33929
engineering project. The tunnel of the project is composed by U-tanks and box culverts. 1 + 568 1 + 638,
1 + 794 1 + 864 are U-tanks. Each tank is 70 m
long and the width of banks is summed to 22 m.
1 + 638 1 + 794 are box culverts.The sum of lengths
of all box culverts reaches 156 m. Equally divided it
into 4 parts, each box culvert is 38.25 m long and the
width of all box culverts is summed to 21.4 m. Reinforced concrete piles with diameter of
100 steel
pipe were taken as support. They are 21 m long with
concrete outside. The concrete piles with
60@30
were used to keep dry from water. The depth of concrete pile is 10 m. They are connected to each other
side by side. The steel in the shape of I is used as
the inside supports. The distance between two supports is 6 m wide. The depth of the foundation pit is
8 m. This foundation pit is level 2 foundation pit. The
constructions 0.000 m level is equal to Huanghai
height +4.125 m.The situation around site area is quiet
complex, especially Jinghang Canal on the west side of
site and ancient municipal heritage Gongcheng Bridge
which is close to the nadir of underground lot, smallest
distance is about 2 m.
GPS HEIGHT APPLICATION
GPS positioning technology has advantages such as

no need of keeping vision between measuring stations, not restrained by weather conditions, able to
measuring the targets 3D displacement and highly
automated. The accuracy of short distance deformation monitoring can reach minor millimeter level[1] ,
thus provides a new method for high-accuracy deformation monitoring of large construction and foundation pit. In Lishui Road projects case the visibility
condition in foundation pit construction site is bad
and most datum marks cant share vision, monitoring marks and datum marks are in different height,
and also there is a across-river benchmark problem.
To solve these problems above, this project take a
monitoring plan using both modern and traditional
measuring technology: using GPS technology to set
up a 3D datum mark network, and using traditional
measuring methods to monitor after the network is
established[2] .
After adjusting GPS measuring results, the outcome
height is geodetic height HGPS relevant to WGS-84
ellipsoid. Since the benchmark height (normal height)
is using in foundation pit engineering application,
the geodetic height HGPS should be transferred into
normal height H0 in this project. The difference
223
between normal and geodetic height is called height

anomaly[3] :
In solving GPS height anomaly, known marks

height anomaly values reliability is crucial to solving
result accuracy. Because of restraint from site condition, it is impossible to have enough GPS marks meet
benchmarks or taking benchmark co-measuring. So
every single marks height anomaly value will make
considerable affect to calculating result accuracy, a
mark with gross error height anomaly value could even
lead to a totally useless result and complete failure.
Thus the initial data should take a gross error test. During the test, the data is normally checked by geometric
conditional closure, like triangle closure in triangle
network or pole condition closure, which monitoring value must meet or by residual from adjustment
error. Since gross error is hard to distinguish from
limited error, this method is hard to discover small
gross error. Also it is hard to find applicable geometric condition closure during GPS height transferring.
To solve this problem, one can pick up some trustful
geometric benchmark spot height and geodetic height
in the GPS network to fit other benchmark height,
or pick some spot separately to processing repetitive
trail calculation, then obtain other measured geometric benchmarks trail height with mathematic model
from fit and using the equation below to obtain fit
residual:
Then we have:

, r11
, r21
, r22
is larger
If one from r10 , r11 , r21 , r22 and r10
than critical value, then we can consider V(n) or V(1)
as anomaly value. After analyzes the sensitivity of
anomaly in r statistics test, Dixon claimed that when

3 n 7, it is better to use r10 or r10
; when 8 n 10,

use r11 or r11
; when 11 n 13, use r21 or r21
; when

14 n 25, use r22 or r22
.
It is natural to use different statistics depending on
different n. When n is small, range estimation has a
better efficiency, but while n become larger, range estimations efficiency decrease accordingly. So when n
is relevant large, use range V(n) V2 or V(n) V(3) to
estimate. Statistics rij or rij s critical value is given
in r(n, ) in reference[4] . is Type 1 probability, also
called significance. Its value usually is 0.05 or 0.01.
When running the test, one can calculate and discriminate from both ends of residual sequence separately, until there is no gross error suspicion in both
ends of the test.
Hi , i is trail height and trail height anomaly, Hi , i

is measured benchmark height and measured height
anomaly. Then, one can use residual to process relevant spots measured benchmark height gross error
test, after carefully analysis of measured value with
gross error, select enough reliable measured value to
run fit again.
HEIGHT ANOMALY GROSS ERROR TEST

METHOD
According to Dixon Test [4] , assume therere a set of

residual V1 , V2 , Vn , sort them from low to high, and
get a sequence like below:
GROSS ERROR TEST EXAMPLE
In Lishui Road underground tunnel foundation pit construction project, the datum marks are the deformation
monitoring datum control system. So they are usually
built in the area outside and far from the construction site to maintain their stability. They should not be
too far though for the consideration of having better
monitoring accuracy and also for our convenience of
work. Our monitoring network is divided into two levels. The first level of monitoring network is composed
by the datum marks and working spots, measured once
a week to maintain its stability. The second level of
the network is set up by working spots and monitoring spots, using stable datum marks to verify working
spots. Six datum marks were set up: four are at the
east bank and other two are at west bank of the ancient
Jinghang Canal. Using GPS to introduce the two datum
marks at west bank of the canal to the canals east in
favor of monitoring network. The datum network is
surveyed four times; following the official construction standards entitled Global Positioning System for
Urban Survey Technique standards CJJ73-97. Three
224
Trimble 4600LS GPS single-frequency receivers were

set up to receive the signal at the same time. The observation time lasted more than 90 min. Information from
510 satellites were efficiently received. The elevation
angle of satellites is 15 degree and a break of 20 sec
was set for every two observations. 12 base lines were
observed and four of them are the repeat ones. Specific
software provided by America supplier was employed
to process the data and to carry out the effective solutions. The maximum of error is about 5 mm while
the minimum is 2 mm. The observation results were
further checked by time synchronized and unsynchronized circle. Datum height network monitoring data is
fit from 4 spots and 16 sets of data of GPS benchmarks
geometric benchmark height residual.
Running Dixon test, first discriminate the largest

residual V(16) , since n = 16, so take r22 and r22
as
statistics.
Using n = 16, = 0.05 as argument, according to

table[4] , r0 (16, 0.05) = 0.507, since r22 < r0 (16, 0.05),
the conclusion is the geometric benchmark height
which V(16) refers to doesnt have gross error. Discrimination of smallest residual V(1) :

As r22
> r0 (16, 0.05), the conclusion is the geometric
benchmark height which V(1) refers to has gross error,
should be eliminated.
After the elimination of residual V(1) , the both ends
test should be run again.

First discriminate the largest residual V(15)
Using n = 16, = 0.05 as argument, according to

table[4] , r0 (15, 0.05) = 0.525, since r22 < r0 (15, 0.05),
the conclusion is the geometric benchmark height

which V(15)
refers to doesnt have gross error. Then

:
test the smallest residualV(1)

Since r22
> r0 (15, 0.05) = 0.525, and r22
< r0
(15, 0.01) = 0.616 it can be concluded that the geo
metric benchmark height which V(1)
refers to doesnt
have gross error.
CONCLUSION
With GPS technologys widely application, people can

simply and efficiently obtain horizontal accuracy of
certain spot on minor millimeter level, but still cant
obtain the spots height on same accuracy level. So
in order to extend GPSs superior ability in surveying 3D displacement, we should put our efforts on
researching how to improve GPS survey accuracy of
vertical displacement, thus it can match with survey
accuracy of horizontal displacement. The reason why
GPS has a low survey accuracy of vertical displacement is that though GPS could provide a high accuracy
geodetic height, the lack of a geodic model with relevant accuracy lead to a serious accuracy decrease
during transferring from GPS geodetic height to normal height. To seek the GPS height anomalys value,
the reliability of known spots height anomaly value is
critical to results accuracy[5] , is the key to improve vertical deformation accuracy. To apply the GPS height
survey in our projects foundation pit monitoring, the
questions below should be considered:
1. Height anomaly is unstable, it maybe smooth in
small range or flat-contour region, where height
datum network of foundation pit monitoring is often
established, thus is easy to seek anomaly value; but
it is very variant in wide range or complex contour
region, possible to occur several value with high
residual. So in order to improve reliability of gross
error detection, when discriminated an anomaly
value, one should analyze carefully before delete it.
2. In calculation of GPS height anomaly, the source of
error is various; it could be surveying error of GPS
geodetic height or GPS geodetic height difference,
or error from geometric benchmark surveying. This
problem directly leads to a difficulty of deciding
error distribution pattern for height anomaly. Since
test method usually run in a certain error distribution pattern, (e.g. Dixon test, requires residual is
random sample from normal distribution) the credibility of using this test to run gross error detection
is decreased in real application.
3. Many factors could affect GPS height component
accuracy. Various measures should be taken to guarantee the accuracy in specific projects. Minimize
the multipath effect when surveying with GPS in
urban area, choosing geodetic type of GPS to monitoring datum during foundation pit construction
period. Experience proved that using these methods
not only avoid the restrictions brought to site conditions from conventional methods, but also improve
working efficiency and assure construction quality.
GPS static relative positioning survey has tremendous practical significance to precise engineering survey. With the reasonable monitoring plan according to
engineering condition and purpose, its accuracy could
225
meet almost every requirements of precise engineering

survey. It also has multiple advantages such as low
cost, high efficiency and a high degree of automation.
The application in Lishui Road foundation pit project
is a useful experience.
REFERENCES
Li, Z.H. & Huang, J.S. GPS measurement and date process.
Wuhan: Wuhan University Press, 2005
Lu, T.D. Zhou, S.J. Guan Y.L. Height anomaly Gross Error
Test and Analysis in GPS Height Conversion. Geotechnical Investigation & Surveying, 2004(4) 5154
Yang, J.T. Jiang Y.X. Zhou J. Analysis on Reliability and
Accuracy of Subsidence Measurement with GPS Technique. Journal of Geodesy and Geodynamics, 2006(1)
7075
Zhang, H. & Gu, J.S. Deformation monitoring and data analysis of foundation in municipal engineering. Journal of
Zhejiang University of Technology, 2003(5) 571574
Zhang, F.G. & Zhang J.Y. Statistical Distribution and Test of
Survey Error. Beijing: China Measurement Press, 1991
226
Study on deformation laws under the construction of semi-reverse method

J. Zhang, G.B. Liu & T. Liu
Department of Geotechnical Engineering Tongji University, Shanghai, P.R. China
ABSTRACT: Taking a 24.09-m-deep foundation pit of Shanghai Metro Line 1 which uses the semi-reverse
construction process of three open excavating-one tunneling as an example, through gathering and analyzing
field monitoring data and making use of forward and back analysis methods, we found out deformation laws of
foundation pit under the construction of semi-reverse method. The implementation results of this project indicated
that the semi-reverse method is an effective way to improve rigidity of the exterior support, control the deformation
of excavation, and ensure safety of the surrounding buildings and pipelines. Meanwhile, the results coincide
essentially with time-space effect. The deformation of the excavation is closely correlative with excavation speed
and exposure time. It provided some useful reference for the design of deep excavation in soft soil.
INTRODUCTION
With the development of urban construction, more

attentions have been paid to the utilization of underground space, and the construction technology level
of foundation engineering has been improved continuously. At the existing construction process of deep
foundation engineering, open excavation method is
the most common construction method at present.
Because it boasts many advantages, such as more
construction operation surface, short period and less
cost. However in the application of some deep excavations which have complicated adjacent environment,
narrow operation space, and complicated geological
conditions, open excavation method would cause great
influence on the traffic flow. At the same time, the
pollution of mud fluid, dust particle, acoustic noise,
and vibration which caused in the construction would
induce discommodity to the residents life. Especially,
open excavation method would go against with deformation control of pits, which would cause perimeter
buildings and structures cracking, and bring great
economic loss or unfavorable social influence. The
complete reverse method has little effect on adjacent
environment, but its speed of excavation is slow, construction technologic process is complicated, and the
cost of pillar piles is high.
Combined with advantages of open excavation
method and complete reverse method, semi-reverse
construction method emerges as the time require, and
has been used more and more widely in Shanghai deep
foundation constructions. Taking a pit of Shanghai
metro line 1, which uses semi-reverse construction

method, as an example, through getting field monitoring data and setting up the finite element model, this
paper has given an evaluation for the characteristics
of semi-reverse method such as construction technology and deformation control laws, kindly expected to
provide with a beneficial reference to those similar
projects in future.
2
2.1
ENGINEERING CASE
General engineering situation
A railway station of Shanghai Rail Transit Line No. 10

(metro line 1) is situated at the intersection of South Xi
Zang Road and Fu Xing Road, and cross transferred
with metro line 8. The geographical position of this
station is shown in Figure 1.
The station of metro line 1 is below that of line 8.The
structure form of this subway station with three floors
is two pillars and three spans, the outside dimension
are 179.2 m (length) 23.8 m (width). And the size
of east and west end well is 27.8 m 16.1 m, which
bottom floor buried depth are 24.06 m, 24.09 m.
According to the requirements of waterproof design
and construction plan, the whole railway station main
body structure is divided into two construction region
with eight parts. The subsection construction drawing
of this station is shown in Figure 2. The west end well
is the first construction part, which requires higher
environment protection. This end well approaches the
227
mud-silt clay layer, muddy clay layer, 11 clay

layer, 12 silt clay layer, 3 silt clay layer, 4 silt
clay layer, 2 fine sand layer. Table 1 shows physical
and mechanical characteristics of different soil layers.

Figure 4 shows Geotechnical section of excavation.
Main hydrology condition of this station is as follows: the shallow groundwater field is phreatic aquifer,
which mainly comes from infiltration of precipitation
and seepage of surface water.The annual average water
stage of Shanghai ranges from 0.50 m0.70 m, and
generally 0.5 m is chosen as design value.
In the report of geological prospecting data, the soil
of 2 fine sand layer is distributed in the site, whose
buried depth is 4446 m and confined water head is
10.511.0 m. Considering the worst factors, when the
pit excavated to 24 m, the coefficient of upheaval in
the bottom of the pit would not meet the requirement
of safety factor, so it should be adopted measurement
for decreasing confined water head.
Figure 1. Geographical position of the station.
Figure 2. Subsection construction of the station.
2.3 Construction procedure
Figure 3. Distribution of monitoring points in west end well.
International Squire (28 floors) and Shen Neng International Building (26 floors), and surrounding with
lots of pipelines. According to the requirements of the
first class environment protection specified for Shanghai subway station, the horizontal deformation of
diaphragm wall should be 1.4H (H is the depth of
excavation), and the maximum settlement of perimeter ground surface should be 1H. The excavation
depth of the west end well is about 24.09 m. It adopts
the underground diaphragm wall with width 1000 mm
and depth 44 m. The brace system applies 1 piece concrete brace of 900 800 and 7 pieces steel tube brace
of
60916. The distribution of monitoring points is
shown in Figure 3.
2.2
Geological condition
Basing on the geological prospecting data, the soils

of engineering site are divided into 9 layers from up
to bottom. They are fill soil layer, silt clay layer,
Considering the actual factors such as construction

period, traffic organization, underground pipelines and
environment protection, the semi-reverse construction
process of three open excavating-one tunneling was
adopted in this project.
The detail of the process is as follows: Firstly excavate the soil to the fifth brace, and then construct the
second median plate between the forth and fifth brace.
With the top reinforced concrete brace, the reversed
median plate and underground diaphragm wall formed
a frame system. While the pavement maintenance of
median plate has been finished, utilize two shield
structure holes of the end well to dig the soil below
the plate until the bottom plate finished.
The semi-reversed construction has brought lots of
inconvenience to excavating and supporting of the pit
under the second median plate. This inconvenience
generally reflects at the narrow perpendicular channel
and the complicated supports installation. Commonly,
installation procedure of supports under the construction of semi-reversed method is that, divide the brace
into several pieces, bring these pieces to the bottom
one after another, and then assemble them together
to the design elevation. For the process as it is mentioned, installing one straight brace generally needs
7hours, and installing one diagonal brace needs 10
hours, which far from the requirement of time-space
effect. Time-space effect requires that the excavation width should be no more than 6m, and the
excavation plus supporting time should be no more
than 24 h (excavation time-16 h; supporting time-8 h).
The deformation should be control ineffectively, if the
pit was not supported within such time. So based on
the engineering traits, the project excavate the soil as
soon as possible while in the open cut period. It uses
open cut method to excavate the soil until 0.5 m below
228
Table 1.
Physical and mechanical characteristics of different soil layers.
Buried
depth
No.
(m)
11
12
3
4
2
2.31
3.41
9.21
19.81
22.91
27.81
42.11
44.71
gravity
peak value of
consolidated
quick shear
Compression
module Es (MPa)
r (kN/m3 )
C (kPa)
( )
Kv
Kh
18.6
17.5
16.7
17.5
18.1
18.2
19.7
19.3
18
11
13
16
15
15
39
0
4.80
3.51
2.27
3.93
4.98
5.08
8.00
15.11
1.97E-92.18E-9
1.17E-91.48E-9
1.77E-9
1.52E-9
1.60E-92.13E-9
9.3E-101.46E-9
2.48E-82.57E-8
2.08E-09
1.33E-09
1.77E-9
1.52E-9
1.87E-09
1.20E-09
2.53E-08
14
16
12
14
17.5
19.5
15.0
31
Figure 4. Geotechnical section of excavation.
the fifth brace (3 m below the second median plate).

When the seventh soil was excavated, disassemble the
supports form the location of the fifth brace and install
them to the location of the seventh brace. So the installation of every brace just needs 0.5 hours, which should
accelerate the construction speed as well as control the
deformation of the pit effectively.
3
FINITE ELEMENT MODELING
Combined with field monitoring data, the finite element software PLAXIS 8.2 (Brink Greve & Vermeer
1998) was used to compute the response of the soil
around the excavation for effective analysis. The problem was simulated assuming plane-strain conditions
and chosen half of the pit as the research subject. The
side boundaries of the mesh (total size 90 m 75 m)
were established beyond the zone of influence of the
Permeability
coefficients (m/s)
settlements induced by the excavation (Caspe 1966;

Hsieh & Ou 1998). The finite element mesh boundary conditions were set using horizontal restraints for
the left and right boundaries and total restraints for the
bottom boundary. The soil stratigraphy was assumed
to be uniform across the site. Those soils with similar
properties would be combined by weighted similarity
method. So six soil layers were compartmentalized for
the calculation simplify.
The soil model used to characterize the clays in the
PLAXIS simulation of the excavation is the hardeningsoil (H-S) model (Schanz et al. 1999).
This effective stress model is formulated within the
framework of elastoplasticity. Plastic strains are calculated assuming multisurface yield criteria. Isotropic
hardening is assumed for both shear and volumetric
strains. The flow rule is nonassociative for frictional
shear hardening and associative for the volumetric cap.
The initial values of the basic H-S input parameters for
the soil layers are referenced as Table 1 and calibrated
by inverse analysis.
The linear spring-layer model is adopted to simulate the braces; the plate element model is adopted
to simulate the underground diaphragm wall, reversed
median plate and bottom plate. Considering the buildings around the pit, 50 kN/m2 overload is applied for
calculation.
Figure 5 shows the calculation model. Table 2 shows
11 calculation phases and the construction stages used
in the finite element simulations. PLAXIS employs
a penalty formulation so that undrained conditions
can be explicitly modeled. Because there was a long
time interval between Phase 6 and Phase 7, the displacements are due to partially drained conditions. So
consolidation should be considered in this stage. Other
stages which not noted as consolidation in Table 2
were modeled as undrained and the excess pore water
pressures were computed relative to some steady-state
value (1m) that changes with dredge line level.
229
Table 2.
Calculation phases and the construction stages used in the finite element simulations.
Identification
Phase no.
Calculation
Stages
Initial equilibrium
Set up diaphragm wall and apply overload
Excavate the first soil and support the first brace
Excavate the second soil and support the second brace
Excavate the third soil and support the third brace
Excavate the forth soil and support the forth brace
Excavate the fifth soil and support the fifth brace
Construct median plate
0
1
2
3
4
5
6
7
Plastic
Plastic
Plastic
Plastic
Plastic
Plastic
Plastic
Plastic
Excavate the sixth soil and support the sixth brace

Excavate the last soil
Construct bottom plate
8
9
10
Plastic
Plastic
Plastic
construction
construction
construction
construction
construction
construction
construction
construction and
consolidation
construction
construction
construction
0
2
4
6
8
10
12
14
depth (m)
16
Figure 5. Calculation model.
18
20
22
24
26
28
support second brace

support third brace
support fourth brace
support fifth brace
construct median plate
support sixth plate
support seventh brace
construct bottom plate
30
COMPARISON OF FIELD DATA WITH

CALCULATION RESULTS OF FINITE
ELEMENT SOFTWARE
32
34
36
38
In order to study the deformation laws under the construction of semi-reversed method, lots of field monitoring data of west end well have been finished from
the second brace has been supported to the roof plate
has been finished. Choose the inclination survey point
CX3 and the settlement points J6-1, J6-2, J6-3, J6-4,
J6-5 which have the same cross section with CX3
as representative points. Combined with calculation
results of finite element software, it could be got
detailed analysis.
4.1
Inclination deformation of underground

diaphragm wall
In the construction process of foundation pit, the displacement curve of inclination point CX3 at different
depth which changed with the working condition is
shown in Figure 6.
From Figure 6, it could be found that the maximum inclination displacement of diaphragm wall is
only 33.69 mm when the bottom plate has been poured,
which is satisfied with the requirement of Class 1
40
42
44
-2 0
8 10 12 14 16 18 20 22 24 26 28 30 32 34 36
accumulative displacements (mm)
Figure 6. Displacement curve of CX3 at different depth

which changed with the working condition.
environment protection. Based on the experiences of

Shanghai underground works these years, with the
similar excavation depth, excavation size, geological
conditions and peripheral circumstance, if the foundation pit adopts open cut method, the deformation value
could not be controlled at so small range.
Calculation results of finite element software and
field data were compared from the time that fifth
braces had been installed, which is shown in Figure 7.
In the figure, dashed line represents the calculation
value, and solid line with circle represents measured
value. Table 3 shows the specific comparative value.
230
The shape of calculation curve was in good agreement with the measured curve, and the maximum value
of calculation deformation was in accordance with
field data while the bottom plate has been constructed.
The results show that finite element method can correctly reflect excavation deformation regularity. It
shows that the diaphragm wall engendered comparative larger deformation within the period from the fifth
brace supported to the median plate constructed. This
is because the discrepancy of the two work conditions
lasts as long as 20 days. Though soils werent excavated, exposure time for the foundation pit with braces
was comparative long. The excavation face is situated
in muddy clay layer which has very strong flow property, and the permeability of the soil is relatively large
(
=1.77 109 m/s). For above reasons the diaphragm
wall engendered larger deformation. In the finite element calculation, consolidation has been considered,
so it could correctly reflect the actual deformation.
0
2
4
10
10
12
12
14
14
16
16
18
18
depth(m)
depth(m)
20
22
24
26
28
20
4.2 Ground settlement
22
24
26
30
32
32
34
38
measured value
calculation value
34
measured value
calculation value
36
36
38
40
40
42
42
44
-2
10
12
44
14
-2
accumulative displacement(mm)
10 12
14
16 18 20 22
24 26 28 30
Support the fifth brace
Construct the median plate
10
10
12
12
14
14
16
16
18
depth(m)
depth(m)
Ground settlement points J6-1, J6-2, J6-3, J6-4, J6-5

are in the same section with the inclination point CX3,
which is distributed with the distance of 3 m for every
point from the edge of pit. Figure 9 shows the field
settlement curves in the process of construction.
From Figure 9, it shows that ground settlement
increased with excavation depth. When the bottom
plate has been finished, the maximum settlement is
only 7.1 mm. The soil presented a little uplift at 6 m
from the edge of excavation. With reference to the
actual engineering project, high pressure jet grouting
was used in the end well for the stability of shield
access to tunnel. It might be the reason that caused the
soil uplifting.
28
30
20
22
24
26
28
18
20
22
24
26
28
30
30
32
measured value
calculation value
34
32
measured value
calculation value
34
36
36
38
38
40
40
42
In order to analyze the relationship between deformation of diaphragm wall and time, we chose the
department of the maximum deformation for filed
data CX3 (CX3-43 point whose depth is 21.5 m) as
a key point. The variations of deformation with time
for this point in the whole excavation construction are
inspected as Figure 8 shows.
From Figure 8, we can find that though the
diaphragm wall engendered large deformation from
the fifth brace supported to the median plate supported, in the process of concrete maintenance, the
deformation stopped to grow and it even had a little
falling, and also when the soil under the reversed media
plate was excavated, the deformation rate is smaller
than that of previous. With the top reinforced concrete
supports, the reversed median plate and underground
diaphragm wall could be formed as a frame system,
which constrained the spreading of soil deformation.
The reduction of deformation rate in this phase has
released the comparatively large deformation, which
engendered as a result of soil creep in forward phase.
This is beneficial for the reduction of foundation
deformation and assurance of pit stability.
42
44
44
-2
10 12 14 16 18 20 22 24 26 28 30 32 34
-2 0
8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40
Support the sixth brace
Support the bottom plate
Figure 7. Measured
displacements.
Table 3.
versus
computed
4.3 Building settlement and pipeline settlement
horizontal
To March 2007, while the roof plate has been finished,

the maximum building settlement was only 3.3 mm,
which is at F05 point. The variations of deformation
Specific values of measured versus computed horizontal displacements.
Project database
Completion time
Measured Value
Calculation value
Maximum value
Depth
Maximum value
Depth
Fifth brace
Median plate
Sixth brace
Seventh brace
Bottom plate
06-11-16
12.33
14
13.19
15
06-12-6
27.94
19
27.32
20
06-12-29
30.25
19
31.05
21
07-1-1
31.85
21
35.35
21
07-1-15
33.69
21.5
37.68
22
231
-5.00
displacement (mm)
0.00
fifth
brace
5.00
10.00
15.00
second
brace
third
brace
median
plate
fourth
brace
20.00
sixth
brace
seventh
brace
bottom
plate
25.00
30.00
35.00
40.00
10-20-06
10-30-06
11-9-06
11-19-06
11-29-06
12-9-06
12-19-06
12-29-06
1-8-07
1-18-07
1-28-07
Figure 8. Variations of displacement with time for CX3-43 in the whole excavation construction.
Figure 9. Field settlements in the process of construction.
Figure 11. Time-history curves of M01, S02 and Y01.
Figure 10. Variations of deformation with time for F05.
with time for F05 are shown in Figure 10. As a result

of high pressure jet grouting, in the former phases of
construction the vertical deformation of F05 presented
an uplifting trend. It was not until January 2007 that
the soil deformation fell back. This was one of the
main reasons for the so small accumulative building
settlement.
The conditions of pipeline settlement were as
follows: the maximum deformation point of gas
pipe was M01, whose accumulative settlement was
8.6 mm; the maximum deformation point of water
supply pipe was S02, whose accumulative settlement was 8.4 mm; the maximum deformation point
of rain pipe is Y01, whose accumulative settlement
was 8.2 mm. Choosing pressure pipe point M01, S02

and non-pressure pipe point Y01 as key point, whose
time-history curves is shown in Figure 11.
The settlement trends of these pipelines were uniform, and these time-history curve shapes were similar
to that of inclination point in Figure 8. From December 6th to December 29th, when the median plate has
been maintained, the deformation value of pipelines
also presented a stable period. In 1969, Peck put forward stratum compensation theory, which indicated
that the shapes and the enclosed area of lateral deformation curves caused by foundation pit excavation are
similar to that of ground settlement curves. From Figure 11, it can be found that this similarity changed
uniformly with time, that it is to say the ground settlement changed with the lateral deformation at any time,
which is favorable for the environment protection.
CONCLUSION
1. Adopting semi-reverse construction method in

metro foundation pit could control the deformation of pit effectively, and decrease the influence
of excavation construction on its surrounding environment. Semi-reverse construction method owns
a deep foundation support technology with practical value and brilliant prospects, which would
232
2.
3.
4.
5.
be further developed and applied in rail transit

construction.
The reversed median plate and underground
diaphragm wall formed a frame system. In the process of median plate maintenance, the deformation
of soil behind retaining wall was stable. When the
plate maintenance finished and the soil excavated,
the deformation rate was smaller than those engineering works which adopted open-cut method of
the same conditions.
In the process of reversed median plate supporting,
a long period was needed for reinforcement assemble and scaffold erection. As a result of soil flow
property, larger deformation may be generated at
this period.
As the excavation is in clay, longer times of construction may result in partial drainage as well.
Consolidation of the soil should be considered in
finite element calculation.
The shapes of lateral deformation curves caused
by foundation pit excavation are similar to that
of ground settlement. This similarity changed uniformly together with time pass. It is conjectured
that the ground settlement changed with the lateral
deformation shape at any moment.
REFERENCES
Brinkgreve, R.B.J. & Vermeer, P.A. 1998. Finite element code
for soil and rock Analysis. PLAXIS 7.0 manual, Balkema,
Rotterdam, The Netherlands.
Liu, J.H. & Hou, X.Y. 1997. Foundation engineering manual.
Beijing: China Construction Industry Press.
Richard, J., Finno, M. & Michele, C. 2005. Supported excavations: observational method and inverse modeling. Journal of geotechnical and geoenvironmental engineering
10.1061:826836.
Zhao, G.W. & Guo, H.B. 2006. Application of semi-reversed
construction method to rail transit construction. Building
Construction 28(10):815818.
233
Comparison of theory and test on excavation causing the variation

of soilmass strength
J. Zhou & J.Q. Wang
Department of Geotechnical Engineering Tongji University, Shanghai, P.R. China
L. Cong
ABSTRACT: In view of the excavation unloading characteristic, the variation of soilmass strength is studied
through the theoretical deduction and the test analysis. Baseed on the Hvorslevs real strength theory, the strength
ratio of the unloading soil and the normal compressed soil considering the pore-water pressure is deduced and
the test simulating excavation is carried out. Through comparing data of the theory and test, the soilmass is
caused to be at the overconsolidated state, and the soil microstructure is damaged, then the soilmass strength is
reduced in the unloading process. The analysis result of theory and test are helpful to the further understanding
of the effect of unloading in excavation on the variation of the soilmass strength, which are very significant for
avoiding project accidents.
INTRODUCTION
With the rapid and remarkable development of city

construction, an increasingly large number of the
exploitation of the underground spaces have emerged,
such as high-story building, the underground market
and underground garage etc., which need to excavate
for building foundation. The excavation, including the
influence of the soilmasss engineering property and
the variation of the environment characteristic, have
been systematically studied by numerous scholars.
Rutledge (1944) summarized the soil sample disturbance to the influence of the unconfined compression
strength and the initial tangential modulus in stress
and strain curve, and the result showed that the initial tangential modulus of the remoulded soil sample
is smaller than the one of the undisturbed value by
about 20%, some were only even 3%4%. On resonant column test, Drnevich & Massarsch (1979)
discovered that even if the soil sample suffered from
the small disturbance, its initial tangential modulus
also obviously reduced. Broms (1980) pointed out
that the soil sample disturbance in the brittle soil
to the stress and strain curves influence was much
bigger than in the plastic soil. Zeng (1995) studied
the subway double lines shield tunnel construction

to the influence of the surface, the buildings and
the underground pipelines, and analyzed the tunnel
interval to the influence of stress and the displacement of surrounding soilmass. Zeng & Pan (1988)
studied stress path to the influence of the undrained
strength in excavation. Wei (1987) has studied the
relation of the excavation unloading and passive soil
pressure.
Several examples of the collapse of foundation pits
in the past had very serious consequences, which urged
the people to study the design and construction of foundation pit deeply. At present the maintenance structure
of foundation pit is designed and calculated by using
the elastic foundation beam law or the elastoplasticity finite element method. The routine-test parameters
generally were adopted as the computation parameters, which had not really considered the excavating
and unloading to the influence of soilmass strength.
The variation of soilmass strength after excavating and
unloading is studied through the analysis of theory and
test in this paper. The result of the study indicates that
the unloading in excavation has influence on the variation of soilmass strength, which can be of some help
to avoid project accidents.
235
2 THEORETICAL DEDUCTION OF SOILMASS

STRENGTH UNDER UNLOADING
cr
After excavation, the surrounding soil can be seen

as the overconsolidated soil layer, and Wei (1987)
deduced the undrained strength of the excavation
unloading soft clay according to the Hvorslev (1960)
real strength theory, which was the same formula that
Mayne (1980) obtained the undrained strength of overconsolidated clay soil according to the statistics of a
large number of test data. After the excavation, in fact,
the effective stress of bottom soil layer of foundation
pit is in unceasingly developing and changing process,
rather than the static overconsolidated state that above
formula derives. In this process, because of unloading, the negative pore water pressure dissipates slowly,
and effective stress decreases gradually, and eventually
stops at the overconsolidated state. According to the
Hvorslev real strength theory, the undrained strength
of the soils after excavation and unloading is deduced.
The Hvorslev strength formula is as follows:
where ce = pe ; In normal consolidated soil, pe is

equal to the current effective stress (drained shear
strength) or the consolidation pressure (undrained
shear strength); In overconsolidated soil, pe is the
consolidation pressure that the test specimen failures porosity ratio corresponds in the normal pressure
dense curve. tge is the increment ratio which the
shearing strength increases along with the effective
stress change when the water content is constant; P is
the effective stress.
According to the real strength theory and the critical
states concept, when soilmass has withstood a simple loadingunloading cycle in stress history, it can
be assumed that the failure point of overconsolidated
soil is coincidence with the failure point of the normal consolidated soil at the same water content when
stress path reaches at critical state line, as shown in
Figure 1 and Figure 2. Then the effective stress of the
overconsolidated soil is as follows:
Based on the confirmation of a large number of

tests, it can be considered approximately that the
shapes of undrained stress path of the normal compacting soil sample is geometrical similarity under
the different consolidation pressure, and the variation
value of effective stress and consolidation pressure are
in proportion (Wei 1987). Therefore their relation can
be proposed from the following equation:
iti
ca
ls
tat
el
in
initial compressed line
p'b
p'a
p'e
normal compacting soil
over-compacting soil
normal compacting soil
p'
pe
pc
Figure 1. The constant consolidation pressure and the

undrained stress path for the normal consolidated soil and
the overconsolidated soil (Wei 1987).
E
C
pe
pc
log p
Figure 2. The relation curves for the normal consolidated

soil and the overconsolidated soil.
The undrained shear strength of overconsolidated

soil sample B is as follows:
Taking the equation (2),(3),(5) into the equation (4),

then:
236
The undrained shear strength of normal consolidated soilmass A in initial consolidation pressure Pc
is as follows:
Figure 3. The diagram of the stress variation for excavation.
In the above deduction process:

Sub The undrained strength of the excavation
unloading soils;
Sua The undrained strength of the normal consolidated soils;
ut The negative pore-water pressure of the unloading soilmass, with time dissipation;
Pt The soilmass consolidation pressure of current
state;
pc The soilmass consolidation pressure under
natural state;
Cc , Cs The compression index and swelling index
of soilmass.
In Figure 2:
On the initial compression line AEC,
The law is reflected in the equation (12) that soilmass strength reduces gradually with the negative
pore-water pressure dissipation after unloading. when
ut = 0, soil is at the complete over-compacting state.
The key that estimates the soilmass strength after
unloading is to determine the parameter i. Because
of the errors of the sampling disturbance and instrumentation equipment, the computed result of the
parameter i which is determined by consolidation tests
result Cc , Cs , is bigger than the real value, needing to
calculate the parameter i by the strength test.
After the synthetical comparison, Mayne (1980)
propose that 1 Cs /Cc takes the statistical average
value 0.64 to be quite reasonable according to the
statistics of a large number of experimental data
and in situ measurement; Furthermore, Zhang & Wei
(1987) have confirmed this viewpoint. Therefore the
parameter i is taken 0.64 in this paper.
3
The equation (12) is the undrained strength ratio

of the excavation unloading soil and the normal
consolidation soil.
EXPERIMENT ANALYSIS OF SOILMASS

STRENGTH AFTER UNLOADING
3.1 Test plans

In the excavation process, as shown in Figure 3, the
soilmass unit A of around foundation pit wall is lateral
237
Table 3. The basic index of mechanical property of Silt clay.
Table 1. The triaxial test plans of constant pressure consolidation.

Consolidation pressure
v = H (kPa)
Variation mode of
the stress
Unloading v unloading
failure
Unloading v loading
failure
v unloading failure
v loading failure
100
200
300
400
I01
I02
I03
I04
J01
J02
J03
J04
K03
L03
Parameters
Values
e
Sr
Ip
IL
1-2
c
52.6%
16.9 KN/m3
1.487
97.3%
23.2
1.231
1.20 MPa1
11 kPa
9.3
Table 4. The negative pore-water pressure data of soil

sample after unloading (kPa).
Table 2. The triaxial test plans of K0 consolidation.

Consolidation pressure
v = H /K0 (kPa)
Variation mode
of the stress
180
240
400
H unloading failure
M01
M02
M03
Serial number
I (J) 01
I (J) 02
I (J) 03
I (J) 04
Negative
pore-water
pressure (kPa)
30.9
34.1
Note: Because of instrument failure, the data of 01 and 02

can not be detected, the latter two data are average value of I,
J group.
unloading, and vertical pressure is nearly invariable;

The soilmass unit B of the lateral and vertical in the bottom of foundation pit simultaneously unloads, and the
vertical stress drops more quickly than lateral stress,
but still retains quite a part of incomplete unloading
stress, therefore the influence of the unloading only
possibly exists in a certain scope below foundation pit
bottom surface. The stage excavation method is generally selected for the foundation pit.The first supporting
structure immediately be taken when the first layer soil
is excavated to reach at the design elevation; Then the
second layer soil is excavated. During the excavation,
the soilmass units are in the process that the soil is
unloading and expanding and the negative pore-water
pressure is dissipating slowly.
According to the unloading characteristic of above
soilmass unit A, B, test plans are designed as shown in
Table 1 and Table 2.
I, J group tests simulate the stress path of the unit
B, and the soil samples are consolidated for 24 hours
under constant pressure; then according to the stress
path v = v /3, v /H = 2, the test specimen are
unloaded simultaneously on the vertical and the lateral, and the value was recorded when the negative
pore-water pressure are stable after unloading; This
process needs for 23 hour from starting unloading to
stability of reading value, then turns on the drain valve,
and makes the negative pore-water pressure dissipation, the soil sample completes consolidation under
the new low stress condition, the consolidation time is
8 hours. Then the I group is loaded to the test specimen
compression failure on the vertical, and the J group
is unloaded to the test specimen extrusion failure on
the vertical. As a reference test, K group and L group

specimen are consolidated in the confining pressure
of setting, without unloading disturbance, and respectively are loaded and unloaded to the test specimen
failure on the vertical.
The unloading stress variation process of unit A is
simulated by M group test. K0 = 0.6, the consolidation
pressure is exerted by staged loading; For avoiding the
accidental failure of the test specimen, the staged loading is divided 10 levels to exert; The axial stress
is exerted in each level loading, at the same time, the
confining pressure H = K0 v is exerting, consolidation time 24 hours. After consolidation completes,
v is maintained invariable, and the test specimen is
compressed to failure by the lateral unloading.
3.2 Test results
The soil samples of tests are the typical soft soil of the
Shanghai area, its basic index of mechanical property
as shown in Table 3.
After unloading, the negative pore-water pressure
is shown in Table 4.
4 THE COMPARISON ANALYSIS BETWEEN
THEORY RESULT AND TEST RESULT
OF THE SOILMASS STRENGTH AFTER
EXCAVATING AND UNLOADING
The theory deduction and the test simulation about
soil strength of the excavation has been discussed.
238
Table 5. The soil strength contrast between theory result

and test result after unloading.
Project
Peak value
(1 3 )max (kPa)
Natural Soil sample Test
Theory Difference
Test
soil
of unloading result
result
value
number sample disturbance Sub /Sua Sub /Sua (%)
I
J
Figure 4. The normalization stress of the unloading soil of

I group.
Figure 5. The normalization stress of the unloading soil of

J group.
Now we will discuss the comparison reuslt of theory

and test.
The stressstrain relation curve of different confining pressure of I, J two groups tests are normalized,
and m = (1 + 22 )/3 = 233 kPa, then the curves can
be drawn as shown in Figure 4 and Figure 5, in which
K Line, L line are the soil stress-strain curves of natural compaction after normalization, and I02, I03, J01,
J02, J04 are the normalizing stress-strain curve of
over-compacting soil after the unloading.
After excavation unloading, the stress value of the
soil sample stress-strain relations curve approaches or
surpasses the stress value of the natural normal compacting soil as shown in Figure 4 and Figure 5; For
eliminating the test error, after unloading, soil sample peak value (1 3 )max can be taken the average
value of each normalized curve peak value, as shown
208.2
205.1
219.0
212.5
1.052
1.040
1.175
1.175
12.3
13.5
in Table 5; For comparing with test result, theoretical

calculation is taken by the formula (12), the parameter Pc = 300 kPa, Pt = 233 kPa, the test results and the
theoretical formula results, are shown in Table 5.
In the Table 5, Sub is the undrained strength of
excavation unloading soil in I, J series tests by
normalization; Sua is the undrained strength of the
normal compressed soils in K, L series tests by
normalization.
In above tests, the influence of negative porewater pressure (Table 4) is considered in soil sample.
According to the computation of the formula (12),
the soilmass strength ratio Sub /Sua is 1.229 and 1.235,
which is higher about 5%6% than the ratio of the
pore-water pressure dissipating completely.
The comparison from Table 5 can be found that theoretical calculation result is bigger about 10% than
test result. Not only The soilmass is caused to be at the
overconsolidated state, but also the soil microstructure
is damaged, and the soilmass strength is reduced in the
unloading process. In this triaxial test, soil samples is
unloaded according to the stress path of the test, soil
stress is redistributed, and consolidated, the overconsolidated soil is formated, soil structure of the original
system is also damaged.
5 THE VARIATION OF SOILMASS STRENGTH

PARAMETERS AFTER EXCAVATING AND
UNLOADING
Based on the stress-strain relation curves of tests,
according to the Mohr-Coulomb criterion, the soilmass strength parameters c$$ of the simulating
excavation unloading and the parameters c$$ of
routine-test are listed in Table 6.
The cohesion, the angle of internal friction that
unloading failure of I and M group obtained are quite
close, their c value is bigger than J group, but the
value is smaller, furthermore, the values of c$$
are obviously different from the result of conventional consolidated quick shear test. Because of lack
239
Table 6. The strength parameter value c$$ of the soilmass

failure.
Test number
A unit of
M group
(unloading
Parameter failure)
B unit of
I group
(unloading
failure)
Cohesion
c (kPa)
Angle of
internal
friction
( )
26.5
24.3
15
13
B unit of The result of

J group
conventional
(loading consolidated
failure)
quick shear
7.36
16
11
9.3
of the sufficient test data, the relationship between the

unloading stress and the strength parameters c$$ can
not be obtained, which needs further test to determine
whether these parameters have the inevitable relation.
6
CONCLUSION
This paper chooses the typical excavation as the test

study object, designs and carries out different stress
paths indoor triaxial tests in I, J, K, L, M group tests.
Some useful conclusions are drawn by analyzing the
influence of excavation on the result of theory and
test, which are very significant for avoiding project
accidents:
1. By the assumption that the soil as overconsolidated
soil with dissipation of the negative pore-water
pressure after unloading, the undrained strength
ratio between the soils of excavation unloading and
the normal consolidated soils is deduced, namely
the formula (12). With the dissipation of soil negative pore-water pressure, the soilmass strength is
reduced. According to the analysis result, the scope
of reduction is not large. Under the above test stress
condition, the strength ratio range of variation is
about 5%6%.
2. The undrained strength ratio Sub /Sua from the tests

is smaller about 10% than undrained strength ratio
from the above theoretical formula computation.
The difference value in the test can be identified as
the result that of unloading disturbance.
3. In the unloading process, the soilmass is caused to
be at the overconsolidated state, the soil microstructure is damaged, and the soilmass strength is
reduced.
4. The total stress strength parameters c$$ obtained
from the different stress path tests are much different from the parameters from the routine-test.
Due to the insufficiency of data, the relationship between the unloading stress and the strength
parameters c$$ can not be obtained, which needs
further test to determine whether these parameters
have the inevitable relation.
REFERENCES
Broms, B.B. 1980. Soil Sampling in Europe: State-of-theArt. Journal of the Geotechnical Engineering Div. 106:
6598.
Drnevich, V.P. & Massarsch, K.R. 1979. Sample Disturbance and Stress Strain Behaviour. ASCE Journal
of the Geotechnical Engineering Division 105(GT 9):
10011016.
Hvorslev, M.J. 1960. Physical component of the shear
strength of saturated clays. Research Conference on Shear
Strength of Cohesive Soils, ASCE: 169274.
Mayne, P.W. 1980. Cam-clay prediction of undrained
strength. Geotech Engrg Div ASCE 106(GT11):
12191242.
Rutledge, P.C. 1944. Relation of undisturbed sampling to
laboratory test. Transactions ASCE, (109): 11551183.
Wei, R.L. 1987. Soft clay strength and deformation. Beijing:
China communications press.
Zeng, X.Q. 1995. Subway project double thread tunnel
parallel advancement interaction and construction
mechanics research. Shanghai: Tongji University doctoral
dissertation.
Zeng, G.X., Pan, Q.Y. & Hu, Y.F. 1988. The Behavior of
Excavation in Soft Clay Ground. Chinese Journal of
Geotechnical Engineering 10: 1322.
240
Theme 2: Construction method, ground treatment,

and conditioning for tunnelling
Ten years of bored tunnels in The Netherlands: Part I, geotechnical issues

K.J. Bakker
COB, Delft University of Technology, Delft, The Netherlands
A. Bezuijen
Deltares, Delft University of Technology, Delft, The Netherlands
ABSTRACT: Ten years have passed since in 1997 for the first time construction of bored tunnels in the
Netherlands soft soil was undertaken. Before that date essentially only immersed tunnels and cut-and-cover
tunnels were constructed in the Netherlands. The first two bored tunnels were Pilot Projects, the 2nd Heinenoord
tunnel and the Botlek Rail tunnel. Since then a series of other bored tunnels has been constructed and some
are still under construction today. At the beginning of this period, amongst others Bakker et al (1997), gave an
overview of the risks related to bored tunnels in soft ground and a plan for research related to the pilot projects
was developed. After that in 1999 the 2nd Heinenoord tunnel opened for the public, the Jointed platform for
Bored tunnelling, in short GPB, was organized, to coordinate further research and monitoring of bored tunnels.
This platform is supervised by the Center for Underground Construction. In this paper a summary is given of
some of the most characteristic observations on these 10 years of underground construction in the Netherlands.
In the first part of this paper the focus is on geotechnical interactions, and stability, whereas part two will focus
more on structural related issues.
INTRODUCTION
In 1992 a fact-finding mission was sent to Japan by

the Dutch government, which reported that it should
be possible to construct bored tunnels in the Dutch
soft soil conditions. Up to that time essentially only
immersed and cut-and-cover tunnels were constructed
in the Netherlands, as boring of tunnels in soft soil conditions, at that time, was considered to be too risk full.
After this conclusion things went quite fast; in 1993
the Dutch minister of Transport and Public works
ordered the undertaking of two pilot projects, the 2nd
Heinenoord Tunnel and the Botlek Rail Tunnel. The
projects were primarily aimed at constructing new
infrastructure and besides that for monitoring and
research in order to advance the development of this
new construction method for the Netherlands. The
projects started in 1997 and 10 years have passed
since then.
At the start of the pilot projects, the difficulties with
respect to constructing bored tunnels in soft soil conditions were evaluated and a plan for monitoring and
research was put forward, see Bakker et al (1997).
Since then, the 2nd Heinenoord tunnel, see Fig. 1, and
a series of other bored tunnels were constructed.
After the completion of the pilot projects a Joint
Platform for Bored tunnels was established (GPB) that
coordinates the monitoring and research at the various
Figure 1. Geological profile at the 2nd Heinenoord tunnel.
other Dutch tunnelling projects. The GPB, an initiative

of the larger clients for underground infrastructure on
the government side, was organised under supervision
of the Netherlands Centre for Underground Construction; COB. The research was organised in such a way
that results of a project would be beneficial for a next
project starting a little later.
Unquestionably a lot has been learned from the performed monitoring and research. The results of this
process have been noticed abroad. In 2005 the Netherlands hosted the fifth International symposium of
TC28 on Underground Construction in Soft Ground.
Researchers and experts from all over the world came
to Amsterdam, to learn about the Dutch observations
on tunnelling and to visit the construction works for
the new North-South city metro system in Amsterdam.
243
Table 1.
Bored tunnels completed after 1997 in the Netherlands.
2nd Heinenoord tunnel

Western Scheldt tunnel
Botlek Rail tunnel
Sophia Rail tunnel
Pannerdensch Canal Rail tunnel
Green Hart tunnel
Road
Road
Rail
Rail
Rail
Rail
Completion
(Year)
Bored length
(m)
Depth
(m)
Outward Diameter
(m)
1999
2003
2004
2005
2005
2006
945 (dual)
6700 (dual)
1835 (dual)
4000 (dual)
1615 (dual)
8.620 (single)
30
60
26
27
25
30
8.3
11.30
9.60
9.60
9.60
14.90
The above event was also the occasion for the

presentation of a book; A decade of progress in tunnelling in the Netherlands by Bezuijen and van Lottum (2006), where this research is described in more
detail.
This paper(s) gives some highlights of the main
research result of the past decade.
2
REVIEW OF THE 1997 SITUATION AND

WHAT CAME AFTER
In the design phase for the 2nd Heinenoord tunnel a

main concern were the soft soil conditions in combination with high water pressures. In general in the
Netherlands the water table is just below the soil surface. Furthermore the 8.3 m outward diameter for this
first large diameter tunnel was a major step forward,
compared to past experience in the Netherlands; experience that was mainly based on constructing bored
tunnels, pipes or conduits up to about 4.0 m diameter.
This gave some concern with respect to the amount of
extrapolation of empiric knowledge.
With respect to the soft-soil conditions, the low stiffness of the Holocene clay and peat layers and the high
groundwater table; nearly up to the soil surface, were
considered a potential hazard and a challenge for bored
tunnels. The soil profile at the 2nd Heinenoord tunnel,
see Fig. 1, is indicative for the heterogeneous character and on occasion sudden changes in underground
soil layering that one might encounter. In addition to
the heterogeneity and the ground water, deformations
due to tunnelling might influence the bearing capacity of any existing piled foundations in the vicinity.
And as the common saying is that the Amsterdam
Forest is underground, one might realize the potential risks involved for the North/South Metro works in
Amsterdam.
Characteristic for a high water table is buoyancy;
the effect that the tunnel might be floating up into the
soft upper layers above the tunnel due to the gradient in
the groundwater pressure. Besides the risk of breaking
up of these soil layers, the rather flexible bedding of
the tunnel and the deformations that this may cause
need to be analysed.
Therefore research was aimed at clarifying the

effects of the soft underground, groundwater effects,
and the effect of tunnelling on piled foundations.
After the successful construction of the two Pilot
projects, a number of other bored tunnelling projects
followed, see Table 1. Mention worth is that the Green
Hart Tunnel holds until recently the record as the
largest diameter bored tunnel in the world.
Still under construction are the tunnels for RandstadRail in Rotterdam, the Hubertus Tunnel for a road
in The Hague and the North/South metro works in
Amsterdam.
With respect to the construction of the North/South
metro works in Amsterdam, the station works have
made quite some progress and the bored tunnel is in
a preparation phase. The elements of the immersed
tunnel; the extension to Amsterdam North under the
river IJ, are waiting for the completion of the immersion trench under the Amsterdam Central Station. For
the bored tunnelling part, the TBM is expected to start
excavation at the end of 2008.
Ten years after the pilot projects, the question arises
whether the observations and related research have
confirmed the above issues to be the critical ones or
that advancing insight may have removed these issues
from the stage and swapped these for other topics
giving more concern.
In this paper some of the characteristic events and
results of this past decade will be described. The choice
for the topics being discussed is influenced by the
projects that both authors were involved with, without intent to minimize the importance of other research
that is not discussed in this paper. Further issues related
to groundwater effects and grouting are described in
more detail in a separate paper in this symposium by
Bezuijen & Talmon (2008).
3
EXPERIENCES WITH BORED TUNNELS IN

THE NETHERLANDS IN THE PAST DECADE
3.1 An instability of the bore front

During the construction of the 2nd Heinenoord Tunnel,
approximately in the middle underneath the river Oude
Maas an instability at the excavation front developed,
244
Figure 2. Support pressures before, during and after the

Blow out at the 2nd Heinenoord tunnel.
Figure 3. Pore water pressure distribution in front of

the TBM.
see Fig. 2; afterward commonly referred to as The

Blow-out (see also Bezuijen & Brassinga, 2001).
Backtracking the situation learned that after that a
pressure drop was observed, in his efforts to restore
frontal support, the machine driver first pumped bentonite to the excavation chamber; considering a deficiency in the bentonite system. When this did not help,
air was pumped to the bore front; not realizing that the
front itself already had collapsed. This collapse created a shortcut between the excavation chamber and
the river. The action of pumping air was noticed by
shipmasters on the river, which reported air bubbles
rising to the water surface, which caused the failure to
be known as the blow-out. In this case the pumping
of air had not been beneficial to the restoration of stability because pressure loss was not the cause but one
of the results of the event.
This frontal stability at the 2nd Heinenoord tunnel has attracted some public attention. Presumably
it is less known that loss of frontal stability has also
occurred since then with some regularity at the other
tunnels under construction in the years after, e.g. during construction of the Sophia Rail Tunnel and the
Green Hart Tunnel, however without much delaying the construction process. At the 2nd Heinenoord
Tunnel, construction work was delayed for several
weeks before the crew succeeded in restoring frontal
stability, filling up the crater in the river bottom with
clay and bringing in swelling clay particles in the
excavation room.
From the evaluation of the monitored pressures
in the excavation room, it appeared that before the
development of the instability, the frontal pressure was
raised above the advised pressure for frontal support;
i.e. at about 470 kPa instead of about 310 kPa. see
Fig. 2 (pressure gauge P15 is in the excavation chamber
at tunnel axis level).
In retrospect it was understood that during standstill, the pressures were raised to get a larger gradient in
the pipes in order to improve the transport of excavated
material; i.e. Kedichem clay that was found in the lower

part of the excavation front and appeared to be difficult
to pump through the hydraulic muck transport system.
The measurements indicate that excavation had
started without releasing pressure to the standard support level during excavation. In that condition instability developed within 15 seconds after that the wheel
started cutting. At stand still, when sufficient time has
passed for a proper vertical cake sealing of bentonite
to build up at the front, a high support pressure is
not much of a problem, as the pressures used are way
below those that might override the passive resistance
at the front. However, as the pressure itself is fluid
pressure, when the cake-sealing is taken away during
excavation, and water can penetrate the front, according to Pascals law for a fluid without shear stresses,
the pressure also works in the vertical direction, and
if this pressure exceeds the vertical soil pressure this
will trigger an uplift and possibly a breaking out of soil
layers, and apparently that is what has happened here.
In their paper on face support Jancsecz and Steiner
(1994), for this reason gave a warning about the limits
to the face support pressure, for situations with little
overburden.
Research learns that for the fine sand that we have
in the Pleistocene sands layers in the Netherlands,
penetration of bentonite in the pores is negligible.
Excavation therefore means removal of the cake sealing; Research by Bezuijen and Brassinga (2001),
indicates that it normally takes about 4 to 5 minutes to
build up a new cake sealing after the excavation wheel
has removed the sealing during excavation. The time
between passings of chisels, in the order of 20 seconds
is too short for that. It must be emphasized that this
effect is not only important for the upper limit to face
support pressures, but may also give a limitation to
the lower limit of the support pressure. A method to
discount for this effect was given by Broere (2001),
see also Fig. 4.
245
Figure 5. Under circumstances the Grout material from the

tail void might flow into the gap behind the tail of the TBM,
giving cause to increased loads.
Figure 4. The effect of removal of the cake sealing during excavation on pore-pressures in the front. The influence
zone for excess pore-pressures may be larger that the zone
normally considered in stability analysis.
The situation of a low soil cover underneath the river

bottom is not the only situation that might be critical to
the above effect, also if the soil cover itself is relatively
light, such as in the case of the thicker layers of peat
overlaying the sand where the Green Hart Tunnel was
excavated, this might lead to a critical situation. A local
failure might be triggered where the generated excess
pore pressure in front of the tunnel face can lift the soft
soil layers.
The knowledge gained with the monitoring of the
2nd Heinenoord tunnel was applied for the Green Hart
tunnel, and may have prevented instabilities at the
bore front at larger scales; see Bezuijen et al. 2001 &
Autuori & Minec (2005).
it was possible to give an accepted interpretation of the

measurement results (Bezuijen & Brassinga, 2004),
and a lot of experience has been gained that has
contributed to a better understanding of the grouting
process and the pressures acting on the tunnel lining.
With these results it was possible to predict grouting
pressures and tunnel loading, see Talmon & Bezuijen
(2005).
Based on various evaluations of the force distribution in the tunnel lining, see amongst others, Bakker
(2000), it came forward that the initial in-situ soil
stresses around the tunnel do not have a dominant
influence on the compressive loading of the tunnel.
Due to the tapering of the TBM and in spite of the tail
void grouting there is a significant release of the radial
stresses around the tunnel, see Fig. 5.
The final loading on the lining relates more to the
effectiveness of the grouting process, the distribution
of the grout openings and the consolidation of the grout
than to the initial in-site soil stresses, see Bezuijen
et al. (2004). Whether this reduction of the in-situ
radial stresses is a lasting effect that will remain for
the full lifespan of the tunnel may depend on the creep
sensitivity of the soil, see also Brinkgreve and Bakker
(2001), and Hashimoto et al. (2008).
3.3 Surface settlements
3.2 Tail void grouting and grouting pressures

To measure the soil pressures on a tunnel lining is difficult. In the start-up phase for the monitoring of the 2nd
Heinenoord Tunnel, a number of international experts
on tunnel engineering advised not to put too much
effort on this topic, as the results would probably be
disappointing. Due to the hardening of the grout, the
period for meaningful pressure measurements would
be short and to prevent bridging effects the size of the
pressure cells would have to be large and therefore
costly.
Still, against this advice, the measurement of
grouting pressures was undertaken, and repeated for
a number of tunnel projects. It appeared that the
interpretation was difficult when the grout has hardened, but for the fresh grout until 17 hour after injection
Hoefsloot et al. (2005), have shown that the application of a stress boundary condition between tunnel
and soil in 3D tunnel analysis has a better corroboration between measurement and calculation of soil
deformations around the tunnel and subsequently of
the loading on the tunnel, than the application of the
so called contraction method.
Although this effect was known in the literature, see
for example Mair and Taylor (1997), for the research
team that worked at the 2nd Heinenoord tunnel the
observation that the numerical predictions of surface
settlements lacked accuracy was disappointing. At the
start the expectations on numerical analysis had been
quite high. Shortly after the first observations were
evaluated it was realized within the team, that it were
only the empirical predictions by Peck (1969) for a
246
Figure
6. Surface
settlements;
measured
back-calculated with different material models.
and
volume loss of about 1% that gave a reasonable corroboration with the measurements. The finite element
calculations, at that time mainly based on an application of the elastic-plastic Mohr-Coulomb model in
combination with a contraction type of modelling for
the tail void loss, predicted a too wide and too shallow
surface settlement.
This disappointing result created a problem for the
intentions to apply 3D numerical analysis in deformation predictions for tunnel projects in urban areas, such
as for the Amsterdam North-South line metro works.
Since then, however, a lot of effort has been put
in the improvement of the numerical prediction of
soil deformations. To begin with it was the project
team for the Amsterdam Metro works, see Van Dijk &
Kaalberg (1998), that gave a first indication for an
improvement, with the proposal to model the stresses
at the tunnel soil interface instead of the deformations.
With the introduction of this grout pressure model the
results improved. Later on, when the physics in the process became better understood, i.e. the importance to
account for the high stiffness of the soil in unloading,
double hardening was introduced with the development of Hardening Soil, as a material model; with this
development, the calculation results largely improved
compared to the measurements, see Fig. 6. The latest
development is the introduction of small strain stiffness in the Hardening Soil Model, see Benz (2006),
which up to now gives the best results, see Mller
(2005).
Theoretically the result might further be improved
introducing anisotropy in the model; such models
are being developed nowadays, e.g. in the framework of European Research; AMGISS, e.g. see
www.ce.strath.ac.uk/amgiss.
4
EVALUATION OF THE LEARNING ISSUES
Nowadays its not the soil deformation during normal excavation process that makes us worry about
surface settlements. With an average tail void loss of
about 0.5% of the diameter or less, the deformation

might only be a problem for situations of underexcavation of buildings or if the structures are located
very close to the excavation track. For tunnels in urban
area, there is more concern with respect to bore-front
stability; especially when the upper stratum of the soil
above the Pleistocene layers, where the tunnels are
usually positioned in, consists of soil with a relative
low density, as in the Netherlands. For these situations
with relatively light upper layers of peat or clays with
organic parts, one has to be very careful controlling
the support pressures during excavation, as on the one
hand there is a lower bound value of the support to
prevent cave in, but on the other hand, the upper limit
triggered by an uplift of light upper layers may also be
not far. This will limit the pressure window to work in.
Front instability has occurred at various tunnelling
projects in the Netherlands. If the tunnel is outside
any urban area this might not give too much problems;
however if the tunnel is underneath a city road system, or close to pile foundations this may cause severe
problems, as instability might cause a sinkhole in the
pavement and foundation settlements.
With respect to the accuracy in the prediction of
soil deformations: Apart from the well known empiric
model of Peck (1969) that predicts the shape of the
settlement trough but not the volume loss, the numerical models have become quite reliable in predicting
surface and subsurface deformations, both vertical
and horizontal. The improvement, mainly achieved
in 2D analysis has opened up the possibility for a
reliable deformation analysis in 3D of tunnelling in
urban areas. For an adequate prediction of deformations it is important to model the grouting pressures as
a boundary condition to the excavation, in combination
with the application of a higher order material model,
that takes into account the small strain deformation
behaviour of sand, see Benz (2006).
Further it is recommended, and planned for, to integrate the Delft Cluster Grout pressure model in the
Plaxis 3D Tunnel software. The latter would contribute
to the applicability of the numerical models as a more
general tool for underground construction. This would
enable a better analysis for the loading on the tail of
the TBM and of the tunnel lining.
Within certain limits some cost saving structural
improvements are expected to be possible and, even
more important, insight is obtained in the mechanism
involved.
5
CONCLUDING REMARKS
Ten Years have passed since the first large diameter

bored tunnelling project in the Netherlands in Soft soil
was undertaken. Since then some world records with
respect to tunnelling have been broken in the Netherlands; i.e. the largest diameter (for the Green Hart
247
Tunnel), the highest outside pressure on a segmental

tunnel (for the Westernscheldt Tunnel), the application
of an Earth Pressure Balance shield in coarse sand,
and the largest length of constructed tube in one day,
(Pannerdensch Canal Tunnel).
Before the underground construction works were
started, and the tunnelling projects were in a pre-design
stage, the softness of the Netherlands underground
attracted a large part of the attention, see Bakker
(1997). In retrospect the influence of a low stiffness
as a source of risk and influence on underground construction was confirmed, but sometimes in a different
perspective, or related to other physical processes than
foreseen.
With respect these new insights the following conclusions were drawn:
1 The low stiffness of the soil may also lead to
increased flexibility of the tunnel tube. The deformation of the tube during hardening of the grout,
and the additional Eigen stresses that this may cause
is still a research topic.
2 For a proper prediction of surface settlements and
soil deformations, it is important to model the
grouting pressures at the interface between soil and
tunnel (or grouting zone). Further to improve the
prediction of the width of the settlement trough,
the use of small strain analysis is advised.
3 During excavation in fine sand, such as the Pleistocene sand layers in the Netherlands, the supporting cake fluid will be removed by the chisels on
the excavation wheel. Therefore, in cases of limited
overburden the upperbound to the support pressure
must be carefully determined to prevent instability
of the overlaying soil.
4 In addition; for the determination of the lower limit
to the support pressure, the increased pore pressures
in the front also needs to be taken into account.
With acknowledgement to the Netherlands Centre for Underground Construction for their consent
to publish about the research they commissioned and
coordinated.
REFERENCES
Autuori, P. & Minec, S. 2005. Large diameter tunnelling under
polders, Proceedings 5th Int. Symposium on Underground
Construction in soft Ground, IS-Amsterdam 2005.
Bakker, K.J., v.d. Berg, P. & Rots, J. 1997. Monitoring soft
soil tunnelling in the Netherlands; an inventory of design
aspects, Proc. ISSMFE, Hamburg.
Bakker, K.J. 2000. Soil Retaining structures, development of
models for structural analysis, Balkema, 2000, Rotterdam.
Benz, T. 2006. Small strain stiffness of soils and its consequences, Doctor Thesis, IGS Universitt Stuttgart.
Bezuijen,A. & Brassinga, H.E. 2001. Blow-out pressure measured in a centrifuge model and in the field. Proc. Int.
Symp. on Modern Tunnelling Science and Techn. Kyoto.
Field measurements of grout pressures during tunnelling
of the Sophia Rail Tunnel. Soils and Foundations vol, 44,
No1, 4150, Feb.
Bezuijen, A. & Talmon, A.M. 2005. Grout properties and
their influence on back fill grouting. Proceeding 5th Int.
Symposium on Underground Construction in soft Ground,
IS-Amsterdam 2005.
Bezuijen, A., Pruiksma, J.P. & van Meerten, H.H. 2001. Pore
pressures in front of tunnel, measurements, calculations
and consequences for stability of tunnel face. Proc. Int.
Symp. on Modern Tunnelling Science and Techn. Kyoto.
Decade of Progress. GeoDelft 19952005.
Bezuijen, A. & Bakker, K.J. 2008. The influence of flow
around a TBM machine. Proceeding 6th Int. Symposium
Bezuijen, A. & Talmon, A.M. 2008. Processes around a
TBM. Proceeding 6th Int. Symposium on Underground
Brinkgreve, R.B.J. & Bakker, K.J. 2001. Time-dependent
behaviour of bore tunnels in soft soil conditions; a
numerical study, Proceedings, ICSMGE Istanbul, Turkey.
Broere, W. 2001. Tunnel face stability & new CPT applications, Delft University Press.
Hashimoto, T., Ye, G.L., Nagaya, J., Konda, T. & Ma, X.F.
2008. Study on earth pressure acting upon shield tunnel
lining in clayey and sandy grounds based on field monitoring. Proceeding 6th Int. Symposium on Underground
Hoefsloot, F.J.M. & Verweij, A. 2005. 4D grouting pressure model PLAXIS. Proceeding 5th Int. Symposium on
Underground Construction in soft Ground, IS-Amsterdam
2005.
Jancsecz, S. & Steiner, W. 1994. Face Support for a large MixShield in heterogeneous ground conditions.Tunnelling 94,
British Tunnelling Association, 57 July 1994.
Mair, R.J. & Taylor, R.N. 1997. Theme Lecture: Bored tunnelling in the urban environment. Proc. 14th ICSMFE,
Balkema, Rotterdam.
Mller, S.C. & Vermeer, P.A. 2005. Prediction of settlements
and structural forces in linings due to tunnelling. Proceeding 5th Int. Symposium on Underground Construction in
soft Ground, IS-Amsterdam 2005.
Pachen, H.M.A., Brassinga, H. & Bezuijen, A. 2005.
Geotechnical centrifuge tests to verify the long-term
behaviour of a bored tunnel, Proc. 5th Int. Symposium on
2005.
Peck, R.B. 1969. Deep excavations and Tunnelling in soft
Ground. Proceedings 7th ICSMFE Mexico.
Talmon, A.M. & Bezuijen, A. 2005. Grouting the tail void of
bored tunnels: the role of hardening and consolidation of
grouts. Proceeding 5th Int. Symposium on Underground
Construction in soft Ground, IS-Amsterdam.
van Dijk, B. & Kaalberg, F.J. 1998. 3-D geotechnical model
for the North/Southline in Amsterdam. In A. Cividini
(Ed.), Application of numerical methods to geotechnical
problems, Wien, 739{750. Springer-Verlag.
248
Ten years of bored tunnels in The Netherlands: Part II structural issues

K.J. Bakker
COB, Delft University of Technology, Delft, The Netherlands
A. Bezuijen
Deltares, Delft University of Technology, Delft, The Netherlands
ABSTRACT: In 1997 for the first time construction of bored tunnels in the Netherlands soft soil was undertaken. Before that date essentially only immersed tunnels and cut-and-cover tunnels were constructed in the
Netherlands. The first two bored tunnels were Pilot Projects, the 2nd Heinenoord tunnel and the Botlek Rail
tunnel. Since then a series of other bored tunnels has been constructed and some are still under construction
today. At the beginning of this period, amongst others Bakker (1997), gave an overview of the risks related to
bored tunnels in soft ground and explained about a plan for research related to the Pilot projects. Ten years have
passed, a lot of monitoring and research has been done. In this paper that is split in two parts a summary is
given of some of the most characteristic observations of these past 10 years of underground construction in the
Netherlands. In this second part, the emphasis will be on structural related issues discussed whereas in part one,
frontal stability, grouting and soil deformations are discussed.
INTRODUCTION
In 1992 the Dutch government sent a fact-finding mission to Japan, to report on the possibility to construct
bored tunnels in the Dutch soft soil conditions. Up to
that time essentially only immersed and cut-and-cover
tunnels were constructed in the Netherlands, as boring of tunnels in soft soil conditions, at that time, was
considered to be too risk full.
After the report, that advised positive, things went
quite fast; in 1993 the Dutch minister of Transport
and Public works ordered the undertaking of two
pilot projects, the 2nd Heinenoord Tunnel and the
Botlek Rail Tunnel. The projects were primarily aimed
at constructing new infrastructure and besides that
for monitoring and research in order to advance the
development of this new construction method for the
Netherlands. The projects started in 1997 and 10 years
have passed since then.
At the start of the pilot projects, the difficulties with
respect to the construction of bored tunnels in soft soil
conditions were evaluated and a plan for monitoring
and research was put forward, see Bakker (1997). Since
then, the 2nd Heinenoord tunnel, and a series of other
bored tunnels were constructed. Unquestionably a lot
has been learned from all the monitoring and research
that was performed.
Figure 1. Trumpet effect in tunnel ring construction.
The results of this process have been noticed abroad.

In 2005 the Netherlands hosted the fifth International
symposium of TC28 on Underground Construction in
Soft Ground. The above event was also the occasion
for the presentation of a book; A decade of progress
in tunnelling in the Netherlands by Bezuijen and van
Lottum (2006), where this research is described in
more detail.
In the present paper some highlights of the main
research result of the past decade will be given. The
paper is split in two parts, where part one includes
some general observations and discusses face support,
249
grouting and surface settlements, whereas part two is

more about structural issues.
2
REVIEW OF THE 1997 SITUATION AND

WHAT CAME AFTER
A main concern with respect to boring tunnels in the

Netherlands were the soft soil conditions; the low stiffness of the Holocene clay and peat layers and the high
groundwater table; nearly up to the soil surface were
considered a potential hazard and a challenge for bored
tunnels.
Furthermore the 8.3 m outward diameter for the
first large diameter tunnel was a major step forward,
compared to past experience; experience that up to that
time was mainly based on constructing bored tunnels,
pipes or conduits up to about 4.0 m diameter.
In addition to that, in general deformations due to
tunnelling might influence the bearing capacity of any
existing piled foundations in the vicinity. And as the
common saying is that the Amsterdam Forest is underground, one might realize the potential risks involved
for the North/South Metro works in Amsterdam.
Characteristic for a high water table are buoyancy
effects. Besides the risk of breaking up of the soft upper
soil layers, the rather flexible bedding of the tunnel and
the deformations that this may cause need to be analysed. Therefore research was aimed at clarifying the
effects of the soft underground, groundwater effects,
and the effect of tunnelling on piled foundations.
Ten years later, the question arises whether the
observations have confirmed the above issues to be the
critical ones. In this paper some of the characteristic
events and results of this past decade will be described.
The choice for the topics being discussed is influenced
by the projects that both authors were involved with,
without intent to minimize the importance of other
research that is not discussed in this paper.
3
3.1
EXPERIENCES WITH BORED TUNNELS

IN THE NETHERLANDS IN THE PAST
DECADE
Figure 2. Damage to the Dowel and notch sockets.
Figure 3. Large-scale tunnel ring testing in the Stevin Laboratories at Delft University (the diameter of the (gray) inner
ring is 8.3 m).
Structural damage
An early experience with the difficulties for bored tunnels in soft ground was the damage to the lining that
occurred during the first 150 metres of construction
of the 2nd Heinenoord Tunnel. On average the damage was too high compared to experiences from abroad
and was considered to be unacceptable. Although, the
integrity of the tunnel was not at stake, there was worry
about the durability of the tunnel and the level of future
maintenance.
Characteristic to the damage was cracking and
spalling of concrete near the dowel and notches see
Fig. 2. Quite often the damage was combined with
differential displacements between subsequent rings

and with leakage. The evaluation report, see Bakker
(2000), attributed the damage to irregularities in the
construction of the lining at the rear of the TBM and
subsequent loading during TBM progress. Further a
correlation of the damage with high jack forces was
observed; these appeared to be necessary to overcome
the friction in this part of the track, which prevented
smooth progress.
With respect to the tunnel ring construction, it is
difficult to erect a stress free perfect circular ring. The
250
zone A
zone B
30
zone C
45
Figure 5. Zones that indicate different effects on piles

foundations.
Figure 4. Test site for the Pile-tunnel interaction test.
ring needs to be built onto the end of a former ring

that already has undergone some loading and deformation from the tail void grouting while it partially
has left the tail of the TBM, see Fig. 3. The further
deformation is characterised by the trumpet shape of
the tubing that develops, see Fig. 1, with the inevitable
related stress development in the lining. The trumpet
shape and the high jacking forces lead to local stress
concentrations and irregular deformations in the lining and occasional to slipping between the different
tunnel elements. The slipping of elements was blamed
to the use of a bituminous material called Kaubit in
the ring joint.
Originally Kaubit strips had been used in the ring
joint. These Kaubit strips, of flexible bituminous like
material, were used to prevent the occurrence of stress
concentrations; so some slipping was meant to occur,
but the dynamic character of the slipping that actually occurred that influenced the final geometry of
the lining and had triggered cracking was unexpected.
Especially the cracking and overloading of the dowel
and notch system was unforeseen.
Failure of the dowel and notch system, see Fig. 2.
led to spalling and in some cases to leakage. In the
cases that leakage was observed this must have been
correlated to damage to the notch at the outer side
of the lining, creating a shortcut to water penetrating
behind the rubber sealing there.
After the main conclusions were drawn, it was
decided to exchange the Kaubit strips for thin plywood
plates. Due to the larger stiffness and shearing resistance, shearing of the concrete elements at large was
further prevented and the damage limited.
Besides this technical measure, the evaluation was
the trigger for the undertaking of fundamental research
into lining design that included large scale physical
testing of tunnel tubing at Delft University see Fig. 4.
In this project that was a combined effort of physical
and numerical testing, the details of assembling tunnel segments into subsequent tunnel rings and these
into a tube were investigated. Amongst others the main
results of the project were reported by Blom (2002),
and Uijl et al (2003). Based on this research it was

decided to omit the dowel and notches for the Green
Hart tunnel; which led to a nearly damage free tunnel
lining.
A different issue, not settled yet, is the durability of plywood and the consequences of wood rot on
the long-term tunnel behavior. An unwanted loss of the
longitudinal pre-stress of a tunnel might influence the
tunnel flexibility and deformations, possibly leading
to leakages. On the other hand, experience learns that
compression largely increases the durability of wood.
The ply wood material is compressed to a strain of more
than 50% during tunnel construction. At such a high
level of straining the wood cells might have collapsed.
3.2 Deformations of the TBM machine during
construction of the Westernscheldt tunnel
During construction of the first tube for the Westernscheldt tunnel, unexpected deformations of the tail of
the TBM were observed; i.e. the air space between
tubing and tail of the TBM narrowed at a certain
stage in an unexpected way. The shape of the observed
deformations did not coincide with the assumed soil
loading and gave the impression that it was a large
deformations effect; i.e. buckling.
At first buckling was not accepted as a cause
because the tapering of the TBM was assumed to
give sufficient stress release to guarantee a sufficient
decrease in isotropic stress. Further a certain bedding
effect was assumed to be always present and the combination would make buckling unlikely. Buckling would
only be plausible for a much higher loading of the tail
of the TBM in combination with the absence of any
bedding reaction.
However, the insights have changed since then. In
general there may be no overall contact between the
soil and the tail of the TBM; when grout is injected in
the tail void, the increased pressure on the soil, compared to the original stress will push the soil from the
TBM and grout will flow between the TBM tail and
the soil, see Fig. 5 in part I of this paper. This means
that the pressures on the TBM tail are higher than
anticipated in the past and there might be no bedding
251
reaction. This could well explain the occurrence of

buckling and the deformations of the TBM tail.
A 1-D calculation model has been developed and is
verified with FEM simulations (Bezuijen & Bakker,
2008). This model shows that also the high stiffness
of soil during unloading, which led to the HS and the
HSsmall material models, made it likely that the common tapering, approximately equal to an equivalent
volume loss of 0.4%, is sufficient to lose the larger part
of the effective radial stresses, which helps to develop
a gap between the tail of TBM and the soil.
The grout pressures exerted on the tail of TBM
might be much higher than the soil stresses, and in
absence of bedding, buckling could well explain for
the deformations.
3.3 The influence of tunnel boring to piled
foundations
Large scale testing of pile foundations was performed
during construction of the 2nd Heinenoord Tunnel.
This was done in order of the Project Bureau of the
Amsterdam North/South metro works to get a better
understanding of the processes.
A trial field with loaded piles and pile configurations was installed in the area near and above the track
of the TBM, see Fig. 4. One of the main concerns was
that due to an increase in pore pressure the effective
stresses around the pile tip might be affected and that
a release in isotropic stresses might trigger a drop in
pile bearing capacity.
However, against this reasoning there is also numerical and analytic evidence, (assuming cylinder symmetric analysis), that indicates that the release in
stresses due to tunnelling is limited to a rather small
plastic zone in the close vicinity of the tunnel lining,
see also Verruijt (1993). The analytical model reveals
that strain as a function of the distance drops as a function of 1/r 2 , which would indicate that the influence
zone would be limited in size. This reasoning in combination with the fact that the strains due to tunnelling in
general are quite small; the largest strains often being
less than 0.5 or 1.0%, makes plastic zones further away
than D/2, measured from the tubing, unlikely. Only
above the tunnel this zone can be larger.
However, reasoning and analysis is one thing; measuring and validation is another; based on the field
measurements and physical model research in Delft
and Cambridge Kaalberg et al. (2005), proposed a zoning as shown in Fig. 5, with the following indicators;
a zone A above the tunnel where the settlement of a
pile is expected to be larger than the soil displacements.
A zone B adjacent to the tunnel, with an inclined
influence line, where the pile will follow the soil deformation at the tip of the pile, and further a zone C,
outside Zone B, where at soil surface level the settlement of the pile will be less than that of the soil surface.
This zoning proposal more or less coincides with the
main results as published by Selemetas (2005) that

were mainly based of physical testing in a geotechnical
centrifuge.
The results published by Kaalberg et al. and others are valid for the average volume loss that can
be expected during tunnelling (0.5 to 1%) Earlier
centrifuge testing by GeoDelft indicated that larger
deformation effects are possible for higher volume
losses (up to 7% was tested). Such volume losses are
well above nowadays practice, but it means that during
a calamity, piles over a larger area may be affected.
3.4 Longitudinal deformations of the tunnel tube
In the paper by Bakker et al. (1997), the development of longitudinal stresses in a tunnel lining due
to irregular bedding in soft soil was mentioned as
an item for research. Irregular bedding that could be
the result of zones with different elasticity or else
due to the stiff foundation of a shaft or bedding
in the deeper Pleistocene layers; especially near the
transition between Holocene and Pleistocene layers.
The measurement of longitudinal stresses in itself has
turned out to be cumbersome. Within the monitoring
scheme for the 2nd Heinenoord a trial measurement
was undertaken. In addition to that measurements
from the Sophia Rail Tunnel were back-analysed with
4D finite element analysis, i.e. (time dependant 3D
analysis), and after that the longitudinal stresses were
also measured during the construction of the Green
Hart Tunnel.
To begin with the latter situation: measurements
were taken with a tubular liquid level devise of the
longitudinal deformations of the tunnel during the
grouting process. From these measurements the observation came forward that the tubing exhibited large
vertical movements, up and down, between 20 to
30 mm during excavation and tail void grouting was
measured, and a total vertical shift of the tubing vertical of about 60 mm at one location (See also Talmon &
Bezuijen, 2008).
This amplitude was surely unexpected and is not
fully accepted yet. Nevertheless it is clear that vertical
deformations do occur in the zone where the grout
material is still fluid, and during excavation and may
lead to an alternating deformation; upwards when the
TBM is excavating and grouting and downwards if the
TBM is at stand still.
With respect to the 3D staged construction analysis of tunnel construction for the Sophia Rail Tunnel,
that was undertaken for the COB F220 committee, a
combined DIANA and PLAXIS 3D analysis was performed, see Hoefsloot et al. (2005). The outcome of
these various analyses more or less coincided; which
might have been expected as the mathematical base
of both models is quite similar, and in general deformations remain small, so the soil reactions will most
probably mainly have been elastic.
252
Figure 7. Longitudinal bending moment due to the construction forces measured and calculated acc. to Bogaards
model.
Figure 6. Conceptual model for the analysis of beam effects

in the tube of a bored tunnel by Boogaards (1999).
The main conclusion with respect to this effect was

that this issue can be properly analysed with a relatively simple model based on the concept of a beam
with an elastic bedding and a series summation, such
as developed by Boogaards & Bakker (1999), see
Fig. 6 and later on applied by Hoefsloot (2002). See
Fig. 7 for a comparison between model outcome and
measurements from the 2nd Heinenoord tunnel.
However, using generally accepted parameters, the
measured deformations are much higher than according to these models. Recently, Talmon et al. (2008)
have presented results that may explain the lower stiffness that are found in the measurements (the lining
stiffness can be lower due to only local contact between
the elements and the soil stiffness reduces due to
unloading of the soil around the tunnel), but these are
not yet generally accepted.
connection at least every 250 m, which is nowadays

more or less the reference situation in the Netherlands.
The task to construct these cross passages is a further technical effort. During the construction of the
Botlek Rail Tunnel a vertical shaft and freezing were
the main construction techniques as the cross passages
could be positioned outside the area under the Oude
Maas River. The positive experience with freezing for
the Botlek Rail Tunnel was helpful in the decision
making for the Westernscheldt Tunnel, but there the
freezing was done from the tunnel tube as the track
underneath the estuary is too long and too deep with
respect of the water table to enable the shaft type
method.
Although the method in itself is costly, its reliability
is an important advantage and therefore it is also used
for the cross passages of the Hubertus Tunnel and is
expected to be used in future projects. For the single
tube Green Hart Tunnel tunnel safety is achieved by
construction of a separation wall with doors.
5
CROSS PASSAGES
The design for the Westernscheldt tunnel in the

Netherlands did trigger a debate on tunnel safety. Some
major accidents with tunnel fires, such as occurred in
the Channel tunnel and at the Mont Blanc tunnel in the
Alps did reveal the vulnerability and relative unsafe situations in tunnels with oncoming traffic or in a single
tube in general.
For the Westernscheldt tunnel, a twin tunnel with
one way traffic per tube, the discussion focussed on
what distance between cross passages would be acceptable to guarantee that escaping people would be able
to find a safe haven by entering the other tube; assuming that the traffic is stopped, by an automatic control
system.The outcome of these safety studies was a cross
EVALUATION OF THE LEARNING

ISSUES
The research on grout pressures, in combination with

the structural research on lining design has gained
us the insight that the lining thickness and the necessary reinforcement are mainly determined by the
loading in the construction phase and to a lesser
degree to the soil pressures. In engineering practice the thickness and reinforcement of the tubing is
mainly determined by the most unfavourable jackforces during TBM excavation in combination with
an unfavourable tail void grouting scenario. Difficulty
with these is that its the contractors prerogative to
decide on the necessary jack-forces that will enable
him to construct the tunnel and also what scenario he
will use for the tail void grouting. This might lead to
253
conservative assumptions in the design office in order

to avoid liabilities if a problem would occur during
construction.
With respect to the generality of this conclusion it
has to be considered that the main observations that
were discussed relate to tunnels that are safely located
in stiff Pleistocene sand layers. We must however consider the possibility of tunnels in softer soil layers
that are more susceptible to consolidation and creep.
Consolidation and creep might counteract the general
tendency of stress release and arching in the soil and
lead to a much higher radial loading. One might think
of a soil pressure on the lining that may be on the
level of the initial soil stresses before tunnel construction; the Ko stress situation or even higher than these
initial stresses. Such a situation was accounted for in
the design for RandstadRail in Rotterdam, where a full
steel lining was chosen for a part of the track where the
tubing mainly rests in the upper much softer Holocene
clay and peat layers, that foreseeable would have an
extra loading on the lining due to consolidation and
creep (Pachen et al. 2005).
However, with respect to lining design, within certain limits some cost saving structural improvements
are expected to be possible and, even more important,
insight is obtained in the mechanisms involved.
6
CONCLUDING REMARKS
Ten Years have passed since the first large diameter

bored tunnelling project in the Netherlands in Soft soil
was undertaken. Before the underground construction
works were started, and the tunnelling projects were
in a pre-design stage, the softness of the Netherlands
underground attracted a large part of the attention, see
Bakker (1997). In retrospect the influence of a low
stiffness as a source of risk and influence on underground construction was confirmed, but sometimes
in a different perspective, or related to other physical
processes than foreseen.
With respect to the new insights gained the following conclusions were drawn:
1 The low stiffness of the ground support may give
rise to increased vulnerability of the lining for jacking forces by theTBM during excavation. Care must
be taken to precise shape of the elements and joints
to prevent too high stresses during assembly.
2 The low stiffness of the soil may also lead to
increased flexibility of the tunnel tube. The deformation of the tube during hardening of the grout,
and the additional Eigen stresses that this may cause
is still a research topic.
3 The stiffer Pleistocene sand layers might not always
be able to follow the tapering of the TBM. It is
expected that this may give rise to gapping behind
the tail of the TBM. If grout penetrates this gap,
this may cause higher loads on the TBM than is

normally assumed.
4 No proof was found that tunnel driving, in normal operation, might give cause to loss of bearing
capacity of piles. Settlements in general are related
to the settlement of the ground and the position of
the pile toe with respect to the zones indicated in
fig. 5.
With acknowledgement to the Netherlands Centre for Underground Construction for their consent
to publish about the research they commissioned and
coordinated, and with thanks to Cees Blom for the use
of some of the figures.
REFERENCES
Bakker, K.J., v.d. Berg, P. & Rots, J. 1997. Monitoring soft
soil tunnelling in the Netherlands; an inventory of design
aspects, Proc. ISSMFE, Hamburg.
Bakker, K.J. 2000. Soil Retaining structures, development of
models for structural analysis, Balkema, 2000, Rotterdam
around a TBM machine. Proceeding 6th Int. Symposium
Decade of Progress. GeoDelft 19952005.
Blom, C.B.M. 2002. Design philosophy of concrete linings
for tunnels in soft Soil, Delft Univ. Press, The Netherlands.
Bogaards, P.J. & Bakker K.J. 1999. Longitudinal bending
in Geomechanics Proc. NUMOG VII: p. 317321.
Hoefsloot, F.J.M. & Bakker, K.J. 2002. Longitudinal effects
bored Hubertus tunnel in The Hague. Proceeding 4th Int.
Symposium on Underground Construction in soft Ground,
IS-Toulouse.
Hoefsloot, F.J.M. & Verweij, A. 2005. 4D grouting pressure model PLAXIS. Proceeding 5th Int. Symposium on
2005.
Kaalberg, F.J., Teunissen, E.A.H., van Tol, A.F. & Bosch, J.W.
2005. Dutch research on the impact of shield tunnelling on
pile foundations. Proceedings of 16th ICSMGE, Osaka.
Pachen, H.M.A., Brassinga, H. & Bezuijen, A. 2005.
Geotechnical centrifuge testst to verify the long-term
behavior of a bored tunnel. Proc. 5th Int. Symposium on
2005.
Selemetas, D., Standing, J.R. & Mair, R.J. 2005. The response
of full-scale piles to tunnelling. Proceeding 5th Int. Symposium on Underground Construction in soft Ground,
IS-Amsterdam 2005.
Talmon, A.M. & Bezuijen, A. 2008. Backfill grouting
research at Groene Hart Tunnel. Proceeding 6th Int. Symposium on Underground Construction in soft ground,
IS-Shanghai 2008.
Uijl, J.A., den, A.H., Vervuurt, J.M., Gijsbers, F.B.J. & van
der Veen, C. 2003. Full scale tests on a segmented tunnel lining. In Proc. ITA World Tunnelling Congress 2003,
Amsterdam, The Netherlands, 1217 April 2003.
Verruijt, A. 1993. Soil Dynamics, Delft University of
Technology.
254
The influence of flow around a TBM machine

A. Bezuijen
GeoDelft, Delft University of Technology, Delft, The Netherlands
K.J. Bakker
Delft University of Technology, Delft, The Netherlands
ABSTRACT: The flow of grout and bentonite around a slurry shield TBM is investigated, using a 1-dimensional
calculation model. The results show that the shield of the TBM is only partly in contact with the surrounding soil
and that for a large extent it is in contact with the liquid grout and, depending on the amount of over cutting, also
in contact with the bentonite that is injected at the tunnel face. These liquids also determine the forces on the
TBM shield. The calculation model presents the pressure distribution on the shield and shows the influence of the
soil properties, the overcutting and the properties of the grout and the bentonite. Assumptions in the model are
checked with 2-D finite element calculations, plane strain and axi-symmetric, that show qualitative agreement,
and a reasonable quantitative agreement. The model is useful to explain the phenomena and as a first estimation
of the influence of flow around the TBM.
INTRODUCTION
Tunnelling through urban areas asks for a minimal

ground loss. Modern TBMs can operate at an average
ground loss of 0.5% or less (Bowers & Mos, 2005). To
be able to calculate the ground loss that can be expected
in this range it is necessary to improve the calculation
methods. One decade ago it was sufficient to calculate
whether the front pressure was sufficient to prevent a
collapse of the bore front and not high enough to get
a blow-out. Nowadays 4-D finite element modelling
is used to estimate the ground loss and the settlement
trough that is the result of it.
In most simulations on TBM tunnelling it is
assumed that the non-excavated soil around the TBM
slides over the shield skin of the TBM (Hoefsloot &
Verweij, 2005; Kasper & Meschke, 2006). As a consequence the tapering of the TBM results directly into
a volume loss and a settlement trough.
However, in reality the calculated settlement trough
is too high, there may be a space between the shield
skin of theTBM and the soil.This is because the cutting
wheel of a TBM is in most cases a bit larger than the
TBM itself. There is only a small difference (a few
centimetres on a TBM with a diameter of 10 metres or
more), but this is significant. It means that for a slurry
shield bentonite can flow from the tunnel face back to
the grout used to fill up the tail void. However, it is
also possible that grout can flow from the tail void to
the tunnel face.
A 1-dimensional model is developed to get information on the order of magnitudes of these flows,
the pressure distribution and the soil deformation. The
model assumes a given pressure distribution at the tunnel face (the face pressure) and at the tail of the TBM
(the grout pressure) and incorporates the yield strength
of both the grout and the bentonite. It takes into account
the overcutting of the cutting wheel and the conical
construction of the TBM (with a bit smaller diameter
at the tail compared to the front). Linear elastic soil
behaviour is assumed.
The paper will briefly describe the model, more
detailed information is presented by Bezuijen (2007),
some example calculations will be presented and the
assumptions will be checked by using 2-D finite
element calculations.
2
GEOMETRY OF A TBM
A TBM shield seems to be a tube with a constant diameter, sketched in a way as in Figure 1. However,
looking more in detail the shield is part of a cone.
The diameter at the front is larger than at the tail. The
difference is only small, normally around 0.4% of the
diameter of the TBM. For a TBM shield with a diameter of 10 m, this means that the diameter of the shield is
4 cm smaller at the tail compared to the diameter just
after the cutter head. This small difference in diameter
allows the TBM to manoeuvre in the soil.
255
Figure 1. Sketch of a slurry shield TBM.
GROUT FLOW AROUND A TBM
3.1 Theory
Assume a TBM with a small change in diameter (a
diameter decrease from front to back). The TBM is
boring in soil that is assumed linear elastic with a shear
modulus G.
Assume that the soil in contact with the shield at
every location of the shield. The tapering of the shield
will lead to a decrease in stresses in the soil around
the TBM going from the tunnel face to the tail. A
simple approach is to neglect the influence of gravity and assume a tunnel that is positioned perfectly
symmetric in the bore hole. In such a situation the
relation between deformation and stress reduction can
be written as (Verruijt, 1993):
Where is the change in pressure, r the change in

radius, r the radius of the tunnel and the grout and G
the shear modulus of the soil around the tunnel.
Calculating the pressure from front to tail of the
TBM, without the influence of grout or bentonite, will
lead to an ongoing pressure reduction.
However, bentonite is injected at the front of the
TBM and grout in the tail. Normally the bentonite pressure is lower than the initial soil pressure and the grout
pressure is higher. Using eq. (1) this means that there
will be a diameter decrease at the front and a diameter
increase at the tail. However, when overcutters are used
it is still possible that there is an opening at the front
of the TBM between the tunnel face and the soil where
bentonite can flow over the shield to the tail. Due to
the high grout pressure it is also likely that grout flows
from the tail over the shield to the front of the TBM.
Somewhere in the middle around the TBM these
flows will meet. Calculating the pressure distribution
is now complicated by the fact that the direction of the
Figure 2. Possible flow directions and sketched pressure

distributions along the TBM. The 3rd situation is worked out
quantitatively, see Figure 3.
flow is of importance. Both grout and bentonite are

usually described as Bingham liquids. This means that
a certain pressure gradient is needed to start a flow.
Therefore there are 3 possible flow situations, see also
Figure 2.
1 Grout flows from the tail to the tunnel face and
bentonite flows from the tunnel face to the tail (this
situation can only occur when there is some volume
loss in the gap between the TBM and the soil for
example due to bleeding of the grout or penetration of the bentonite into the soil. For this situation
the lowest pressure will be present there where the
bentonite and the grout meet.
2 Bentonite is flowing backwards to the tail and pushing the grout out of the gap between the TBM and
the soil. The pressure will be the highest at the tunnel face and will decrease when going to the tail.
This cannot be a continuous situation, but can occur
temporally.
3 Grout is flowing to the tunnel face and pushes the
bentonite to the tunnel face. The pressure is highest
at the tail close to the injection points of the grout
and will decrease going to tunnel face.
The model developed can in principle cope with all
these 3 options. However, for simplicity in this paper
256
we will only deal with the 3rd option because this is

the most realistic situation for most of the drilling process. As the machine moves forward, a forward flow of
material is needed to fill the annulus that is created by
the TBM. It is therefore assumed that the grout flows
along the TBM shield and pushes the bentonite away.
Considering the advancing TBM during drilling, this
can be a stable situation. Both the bentonite and the
grout can be described as Bingham liquids. For the
low flow velocities during tunnelling the yield stress
of the liquid will be determining the pressure drop and
the viscous forces can be neglected. Assuming that the
flow induced friction will develop between the soil and
the grout (in the 3rd option the grout and bentonite will
not move with respect to the TBM in a stable situation),
the pressure drop can be written as:
Where P is the change in pressure due to the flow,

x a length increment along the TBM, s the gap width
between the tunnel and the soil and the shear stress
of the grout around the TBM.
The following calculation procedure is used. The
soil around the tunnel is assumed to behave as independent slices with a thicknessx. Knowing the geometry
of the tunnel, the grout pressure at the tail and the bentonite pressure at the tunnel face, the soil pressure and
the elastic properties of the soil, the gap width at the
face and tail of the TBM can be calculated using Equation (1). That gap width can be used to calculate the
pressure drop due to the flow of the grout to the front
of the TBM over the distance x with Equation (2)
and the pressure increase that will occur in the bentonite when it is pushed to the front. The resulting grout
pressure and bentonite pressure is calculated independently. As long as at a certain location the calculated
grout pressure is higher than the calculated bentonite
pressure, it is assumed that the grout pushes the bentonite in the direction of the tunnel face and there will
be grout at that location. In the case that the bentonite
pressure is higher, there will be bentonite.
The result of such a calculation is shown in Figure 3.
The upper plot of this figure shows the pressures and
gap widths that would occur if there were only grout
(G) or only bentonite (B) in dots and in lines the combined gap width and pressure. In this calculation the
grout penetrates between the shield and the soil up
to 1.8 m from the tail. The remaining part of the gap
between the TBM and the soil is filled with bentonite.
The amount of penetration varies with the pressures
and shear strength that are chosen. The parameters
used in this calculation are shown in Table 1. Although
there is an open connection between the face and tail
of the TBM (See the lower plot of Figure 3, the gap
width of the combined calculation is never zero), there
Figure 3. Pressures and gap width along a TBM. Grout pressures and bentonite pressures. Parameters see Table 1. Plots
show pressures and gap width for the bentonite and grout
pressure separately and the combined result.
Table 1. Input parameters used in calculation with bentonite
and over cutting.
length TBM shield
Diameter
diameter reduction
over cutting
asymmetric (1) or
symmetric (2)
grain stress
grout pressure
pore pressure
pressure on tunnel face
shear modulus (G)
shear strength grout, 1 sided
friction assumed
shear strength bentonite 2
sided friction assumed
5
10
0.2
0.015
2
m
m
%
m
150
400
200
250
90
1.6
kPa
kPa
kPa
kPa
MPa
kPa
0.08
kPa
is still a stable boundary between the grout en the bentonite. Since both liquids behave as Bingham liquids,
there can be some decrease in the grout pressure and
still the position of the grout-bentonite front will be
the same.
257
According to the results, the tail of the TBM is never

in contact with the soil. In reality there can be some
contact due to forces not taken into account in this
calculation (vertical loading from the lining, moments
induced by the hydraulic jacks). However, it is clear
that tapering of the shield does not necessarily lead to
a certain volume loss, as was assumed by Kaspar &
Meske (2006). The calculation results presented here
are also in line with the TBM data, which normally
show that more grout have to be pumped into the tail
void than corresponding to the original tail void. The
grout volume that flows into the gap between the soil
and the TBM is not an extra volume, because it is
a constant volume during tunnelling, but the grout
volume increases because the grout pressure makes
a wider gap.
In the Western Scheldt Tunnel project a soil sample
was taken trough the tail of the shield. In that case no
grout was found between the shield and the subsoil,
only some bentonite in the sand (Thewes, 2007). This
indicates that in reality the soil deformation is not as
uniform as suggested here. At some locations the tunnel will remain in contact with the TBM. This will also
be the result of the numerical calculations presented
in the next session.
Figure 4. Mesh used in 2-D simulation. The diameter of the

tunnel is 10 m.
Table 2. Input parameters for the Finite Element analysis.
Parameters give (from top to bottom) the wet volumetric weight, elasticity parameters (Brinkgeve et al., 2006),
cohesion, friction and dilatancy angle.
sat
ur
Eref
oed
Eref
50
Eref
ur
Gref
0
0.7

c
NUMERICAL SIMULATIONS
Some of the assumptions in the calculation model have

been tested by means of numerical calculations, using
the Plaxis programme.
The assumptions tested are the axial-symmetric
deformation of the soil that is assumed and the influence of the non linear behaviour of the sand. Therefore
a 2-dimensional plane strain simulation is performed
using the mesh shown in Figure 4. The simulation is
performed for soil conditions that are common in the
western part of the Netherlands: A Holocene top layer
or 10 m and a tunnel with a diameter of 10 m embedded
in Pleistocene sand. The small strain hardening soil
model was used in the simulation (Brinkgeve et al.,
2006). The parameters used are presented in Table 2.
In the simulation it is assumed that the pressure
on top of the tunnel is 350 kPa at the beginning and
the pressure is increased up to 400 kPa to simulate
the increase in pressure due to grouting behind the
TBM. In a second calculation the stress is first reduced
to 320 kPa and increased to 400 kPa after that. This
simulates the unloading in the soil that can occur due
to the overcutting and/or lower bentonite pressures.
The results will show to what extent a linear elastic
soil behaviour can be assumed.
The pressure increase with depth in the grout is
assumed to be 18 kPa/m. The calculated displacements
around the tunnel opening due to the pressure increase
are shown in Figure 5 for both loading situations. With
kN/m3
MPa
MPa
Mpa
Mpa
kPa
Holocene
Pleistocene Sand
Grout
18
0.35
10
10
5
20
20
0.2
180
180
720
675
1.00E-04
1
32
2
pm
0.499
pm
0.45
15
0
Figure 5. Deformation pattern around tunnel due to a pressure increase of 50 kPa starting at a pressure of 350 kPa (left),
resulting in a maximum displacement of 22 mm. Right the
same calculations but first a stress reduction to 320 kPa was
applied, resulting in a max deformation of 25 mm.
the parameters used in the simulation, the maximum

displacements are in the same order as the displacements calculated with the analytic model (27 mm),
but this depends on the choice of the parameters.
258
It is clear that the displacement is not symmetrical

around the tunnel axis. The soil will predominantly
move upwards. For the situation where there was first
unloading the deformation pattern is a bit more egg
shaped (see Figure 5).
The deformation pattern is therefore not completely
axial-symmetric as is assumed in the analytical calculations. Furthermore, in reality the TBM can rotate
with a horizontal rotation axis perpendicular to the
axis of the TBM in the just created opening. This can
also lead to an uneven distribution of the gap around
the TBM. Therefore from this calculation it cannot
be concluded what will be the gap width distribution
between the TBM and the soil around the tunnel at different locations, what can be concluded is that the gap
width shall not be evenly distributed along the TBM
circumference. Simulations of the deformation of the
soil movements for a tunnel at a larger depth resulted
in a more symmetric deformation pattern.
5
DISCUSSION
The results of the numerical calculations show qualitative agreement with the results of the analytical model.
However, the symmetric deformation that is assumed
in the numerical model is not found in the numerical
simulations, but looking at the results it is reasonable
to assume such a symmetric deformation as a first
assumption. The order of magnitude of the deformation is the same, but this is to a large extend a matter
of tuning the parameters. The parameters used in the
small strain model lead to a stiffer behaviour of the
sand than is normally assumed. When in a case study
reliable soil data are available, both the elastic parameters and the parameters from the small strain model
can be determined.
The influence of unloading before loading seems to
be limited and therefore an elastic model can be used.
Based on analytical calculations of Wang & Dussault
(1994) it can be expected that the influence of unloading becomes more important when the pressures are
decreased further.
The model, as it is now, is suitable to get a qualitative
indication on what is the influence of bentonite and
grout flow. In this way it is a step forward compared to
models that simply assume that the soil flows over the
TBM and that predict a settlement trough that is too
large. For quantitative results and to see what parts of
theTBM are in contact with soil and what are in contact
with the grout or bentonite, more sophisticated models
are needed then the one described here.
6
CONCLUSIONS
flow around a TBM on the pressure distribution in the

soil around the TBM and the influence on the settlement trough. Calculations with this model show that
the influence of the flow alaong the TBM can be significant, leading to calculated volume losses that are
more in agreement with the measurements compared
to results of calculations that do not take into account
the flow around the TBM. A consequence is that the
TBM is (partly) not in contact with the soil. At high
soil pressures and therefore also high grout pressures
this may increase the risk of deformation of the tail
shield.
Comparing the results of the analytical model with
the results of numerical calculations showed that some
of the assumptions in the model (symmetric deformation, linear elastic soil behaviour) are a somewhat
crude representation of reality. This means that it is
worthwhile to include the principles described in the
paper in more sophisticated numerical models. The
model as described here can be used for a first estimate on what is the influence of bentonite and grout
flow around a TBM on the settlement trough.
REFERENCES
Bezuijen A. 2007. Bentonite and grout flow around a TBM.
Proc. ITA 2007, Prague.
a bored tunnel. Proc ICOF 2003 Dundee.
Field measurements of grout pressures during tunneling
of the Sophia Rail tunnel. Soils and Foundations 44(1):
4150.
Bowers, K.H. & Moss, N.A. 2005. Setllement due to tunnelling on hte CTRL London tunnels. Proc. 5th. Int. symposium on Geotch. Aspects of Underground Construction
in Soft Ground, Amsterdam.
Brinkgreve, R.B.J., Bakker, K.J. & Bonnier, P.G. 2006. The
relevance of small-strain soil stiffness in numerical simulation of excavation and tunnelling projects. Proceedings
Numog XII, Graz, Austria.
Hoefsloot, F.J.M. & Verweij, A. 2005. 4D Grouting pressure
model PLAXIS. Proc. 5th. Int. symposium on Geotch.
Aspects of Underground Construction in Soft Ground,
Amsterdam, pp 529534.
Kasper, T. & Meschke, G. 2006. On the influence of face
pressure, grouting pressure and TBM design in soft
ground tunnelling. Tunnelling and Underground Space
Technology 21 160171.
Thewes, M. 2007. Private communication.
Verruijt, A. 1993. Soil Dynamics. Delft University of Technology, b28.
Wang, Y. & Dusseault, M.B. 1994, Stresses around a circular
opening in an elastoplastic porous Medium Subjected to
repeated hydraulic loading. Int. J. Rock Mech. Min. Sci. &
Geomech abstr. 31(6): 597616.
A one dimensional analytical model has been developed to describe the influence of bentonite and grout
259
Mechanisms that determine between fracture and compaction

grouting in sand
A. Bezuijen & A.F. van Tol
Delft University of Technology/Deltares, Delft, The Netherlands
M.P.M. Sanders
Present affiliation: Royal Haskoning, The Netherlands
ABSTRACT: Laboratory tests have shown that injection of grout in sand can lead to different shapes of the
grout in the sand. At low water cement ratios (12) there is usually compaction grouting, which leads to a
spherical shape of the grout. At higher water cement ratios (520) fracturing of the sand will occur, leading to
rather thin grout structures. In field observations this difference is less obvious. It is assumed that in most cases
fractures occur, although it is hardly ever possible to examine what is made in the soil. This paper describes
conceptual models and calculation models that explain, at least qualitatively the differences in the behavior of
the grout and discusses a possible reason for the difference between the model tests and the field tests.
INTRODUCTION
Compensation grouting has been successfully applied

in several projects to prevent or to compensate for surface settlements induced by for example tunnelling
(Mair & Hight, 1994; Chiriotti et al. 2005; Christiaens
et al. 2005). Compensation grouting uses hydraulic
fracturing to get a heave that can compensate the settlement or compaction to densify the soil with limited
heave. In the latter situation the compaction lead to
improvement of the soil characteristics.
Laboratory tests have been performed to investigate
the mechanisms that are of importance for compensation grouting in clay (Jaworski et al. 1981, Mori &
Tamura 1987, Andersen et al. 1994, Chin & Bolton
1999, Au, 2001). Comparable tests have been performed for compensation grouting on sand (Chang,
2004, Kleinlugtenbelt et al. 2006, Bezuijen et al. 2007
and Gafar & Soga, 2006). These tests showed that
compensation grouting by fracturing is not so straight
forward in sand. After quite a number of tests with a
water cement ratio of 12 it appeared that there were
hardly any fractures in the sand, but a more or less
spherical shape of grout is formed in the sand, see
Figure 1. Fractures could be obtained when grout with
a higher water cement ratio (520) is used (Bezuijen
et al. 2007).
Similar results were found in at least 3 different
laboratories. Different results were, however reported
from the field (Grotenhuis, 2004) based on measured
Figure 1. Results from Cambridge University (left) and

GeoDelft that show more a spherical shape than fractures.
surface displacements. Normally it is not possible to

investigate the fractures created, but this was possible
during the construction of the station for the high speed
train in Antwerp and during the construction of the
cross-passages at the Hubertus Tunnel in The Hague,
the Netherlands.
In Antwerp compensation grouting was applied to
prevent settlements on the existing station (Christiaens
et al. 2005). The resulting fractures were found in a
soil layer with a lot of shells. A clear fracture was
found between the shells. Here it is possible that the
shells have influenced the fracture. The Hubertus Tunnel case was an unintended fracture that occurred when
the lances for freezing were brought in and the grout
pressure used to install these lances was large enough
to create fractures.
261
The paper will describe the Hubertus Tunnel case

more in detail, since this is one of the very few cases
where fractures in relatively homogeneous sand at
11 m depth and 5 m below the waterline could be
studied and describes the principle and results of calculation models that are developed to describe various
parts of the grouting process. Possible causes of the
differences between the field test and the model tests
are: 1) the preconditioning has an influence. 2) the
in-homogeneities in real soil serve as a trigger for
the start of the process. 3) The installation of the TAMs
(TAM stand for Tube-a-Manchette, a tube with injection openings covered by a rubber Manchette through
which the grout can be injected) which causes unloading of the soil has an influence. In this paper we
will elaborate the 3rd possibility and mention some
aspects of the 2nd. The first possibility is not dealt
with in this paper.
Results of the model tests itself will be described
and analyzed in another paper on this conference
(Gafar at al., 2008).
2 THE HUBERTUS TUNNEL

The Hubertus Tunnel is a double track road tunnel that
is constructed in dune sand. The tunnels have a diameter of 10 m. Five connections were made between
the two tunnels. Soil freezing was used to make these
connections. Lances were constructed from the tunnel
lining into the sand to freeze the soil. A connection
between the two tunnels was made through the linings
and the concrete connection structure was made. For
a cross-section see Figure 2.
The tunnel crown is placed 11 m below the soil surface and the phreatic level is 5 m below the surface.The
vertical pressure at the top of the tunnel is estimated
to be around 140 kPa.
Grout was used to install the lances for freezing.
During the installation of these lances for the first connection, the pressure was increased so far that the grout
created hydraulic fractures. These fractures became
visible when the frozen soil was removed.
Figure 3 shows an example of fractures in the sand
around the tunnel and between the tail void grout and
the sand.
The picture is taken standing in the newly created
opening, from left to right you see the opening, the
tunnel lining, the tail void grout, grout from a lance
that penetrated between the tail void grout and the sand
and the frozen sand with a fracture in the upper part
of the sand. There were also fractures that were only
visible in the sand.
The normal pressure to inject the grout is 3 bar.
However, according to the employees of the contractor on the site the pressure could have been higher
during this injection because there were doubts on the
Figure 2. Sketch of a connection between the tunnels made

by freezing.
pressure gauge. The grout had a water-cement ratio

(W/C) of approximately 1 and 3% of bentonite was
used in the mixture.
3
COMPARISON WITH LABORATORY TESTS
The laboratory tests performed at GeoDelft, as

described by Bezuijen et al. (2007) and Gafar et al.
(2008) have been performed in a cylindrical tank with
a diameter of 0.9 m and a height of the sand sample
of 0.85 m. The sand was pressurized to get a vertical
effective stress of 100 kPa. Different grout mixtures
were injected and different injection procedures were
used. Most mixtures injected had 5% or 7% bentonite. A fracture like behaviour only occurred when
the W/C ratio in the mixture was 2 or higher. Lower
W/C ratios would not create fractures, but lead to the
shapes shown in Figure 1 (right picture). The maximum injection pressures varied in most cases between
20 and 30 bar.
Comparing the main results of the model tests with
that of the field observations, there seems to be a discrepancy. The grout that does not lead to fractures in
the model, does result in fractures in the field. The
injection pressure in the field situation described is
not known, but it is unlikely that where the normal
262
Figure 4. Sketch with possible deformation modes of the

injection hole. In reality there will be more grains around the
injection hole. The figure just shows the principle.
Figure 3. Fractures found during the construction of the first

connection between the tunnels of the Hubertus tunnel.
injection pressure is 3 bar, it was 28 bar during the

construction of the first cross-section (this would be
the injection pressure based on the laboratory tests
and extrapolating the results of these tests to 140 kPa
vertical pressure).
A possible reason for the discrepancy between the
model tests and the field behaviour can be the soil
boundary condition, as will be explained in the next
sections.
4
4.1
RELEVANT GROUT PROPERTIES

Plastering
Bezuijen & Van Tol (2007) have described how plastering of the grout influences the fracturing behaviour.
Fracturing occurs because the radial stress in the
sand around a fracture is higher than the angular
stress. Using the Mohr-Coulomb criterion, the relation
between radial stress and angular stress perpendicular
on the radial stress can be written as:
a friction angle of 35 degrees, the tangential stress is

only 0.27 times the radial stress.
Looking at the scale of the grains, sand is not
homogeneous and isotropic, as assumed in continuum
mechanics. At the boundary of the opening there will
be some grains that are in closer contact and some
between which there is some space, see Figure 4.
Such space can be sufficient to have the fluid pressure not only in radial but also in tangential direction.
The fluid pressure is with perfect plastering equal to
the radial stress and much higher than the tangential
stress and therefore this will lead to an opening of
the space between the grains and a fracture can occur.
Some in-homogeneity in the soil as is often present
in a field situation (more than in the model test) will
facilitate the fracture formation, because weak sections in the soil will deform. If there is a beginning of
a fracture then the fluid pressure will penetrate further
in the sand and the fracture will easily grow further.
A fracture will stop to propagate when the pressure in
the fracture tip drops due to friction losses, leak-off or
increase of the volume with injection fluid.
Plastering and the formation of a filter cake in
the injection hole will hamper fracturing of the sand,
because now possible irregularities at the boundaries
of the injection hole will be filled with plaster see
Figure 5. This plaster also has certain strength and
therefore prevents the fluid pressure from penetrating
into the space between the grains.
As is also mentioned in Bezuijen en Van Tol (2007),
plastering is caused both by consolidation of the grout
and by leak-off of grout in the sand while larger particles in the grout are filtered by the sand, remain at
in the injection hole and cause a plastering layer at the
boundary between the injection liquid and the sand.
With a constant injection pressure, the consolidation of the grout mixture, without leak-off, can be
approximated by formula (Bezuijen et al. 2007):
Where r is the radial stress, the tangential stress

and the friction angle of the sand. Taking for example
263
Figure 6. Sketch of casing and TAM during installation.
5
Figure 5. Influence of plastering.
Where s1 is the thickness of the consolidated layer (the

plastering layer), ni the initial porosity, ne the porosity
after consolidation, k the permeability of the consolidated layer, the difference in piezometric head over
the column and t the time.
This formula implies that the thickness of the plastering layer increases with the square root of the
applied pressure, a lower pressure will lead to a thinner
plastering layer.
4.2 Leak-off
Assuming that the flow during leak-off can be
described as a Bingham liquid flow, the relation
between the difference in piezometric head and the
thickness of the plastering layer of courser materials
can be written as (for v > 0, flow into the sand):
Where is the difference in piezometric head, s1

the thickness of the grout cake (the plastering layer),
s2 the distance the liquid has penetrated into the sand,
k1 and k2 the permeability of the cake and the sand
respectively for the penetrating liquid, v the velocity,
1 and 2 factors for the cake and the sand respectively
that determine the drop in piezometric head caused by
the yield stress in a Bingham liquid and R the flow
resistance from the soil around the fracture. When
this equation is dominant, the thickness of the plastering will be proportional with the applied difference
in piezometric head.
Experimental research by Sanders (2007) has
shown that consolidation determines the thickness of
the plastering layer for grout with a low W/C ratio, leak
off is the dominant mechanism for the thickness of this
plastering layer at W/C ratios of 5 and more. Normally
a filter cake caused by consolidation of leakoff is made
of finer particles than the sand of the subsoil. Due to
this leak off will stop when there is some thickness of
the filter cake.
INFLUENCE OF SOIL DEFORMATION
In the experiments preformed in Waterloo, Cambridge

and Delft, the sand around the injection point is reasonable uniform and the vertical confinement pressure
is constant over the soil sample. Injection of grout will
lead to cavity expansion until a fracture occurs.
A different pressure distribution is present around
the injection points at the Hubertus Tunnel and in general when compensation grouting is applied in the
field. At the Hubertus tunnel the freezing lances were
applied between the two tunnels. Consolidation of the
tail void grout leads to a stress reduction in the vicinity of the tunnel (Bezuijen & Talmon 2003). Arching
between the 2 tunnels can lead to a further reduction
of the stresses. Also the placement of lances, or in case
of compensation grouting of TAMs, leads to a reduction of stresses. The TAMs are placed using a casing;
see Figure 6 for a sketch. Figure 6 shows the situation
where the TAM is placed by displacement of the soil.
Another option is that the bentonite is used to remove
the soil. The latter option leads to more unloading of
the soil. Whatever option is chosen, when the casing
is removed there will be some unloading of the soil
although grout is injected to stabilize the bore hole, as
appears from (minimal) settlements measured during
the installation of the TAMs. (Kleinlugtenbelt, 2006).
The situation that occurred during compensation
grouting can be compared with the situation of a cavity
under repeated loading. Such a situation is analyzed by
Wang & Dusseault (1994). They used cavity expansion
theory to calculate the stress distribution in the soil
around a bore hole during unloading and re-loading. In
the case of compensation grouting, the removal of the
casing will result in unloading and the reloading will
occur during cavity expansion. Cavity contraction and
expansion theory cannot describe when a fracture will
start, but it can be used to calculate what pressures will
lead to a plastic zone in the soil. Some plastic deformation in the sand is necessary before a fracture can
occur. Plastic deformation will lead to dilatancy in the
sand, which results in more space between the grains.
This makes it easier for the grout to separate the grains
further according to the mechanism described in Section 4.1 leading to a fracture, because there are locally
zones with a higher permeability.
264
Parts of Wang & Dusseaults analyses will be

presented here in a shortened and slightly adapted
version to allow calculation of the influence of previous unloading on the loading pressure at which plastic
deformation occurs. It is assumed that after installation of a TAM there is a pressure release due to the
removal of the casing that leads to an active failure
of the soil around the hole. This means that the radial
stress, r , is smaller than the tangential stress, . After
installation, the radial pressure is increased due to the
fracturing process and passive yield will occur. In such
a situation r , is larger than the tangential stress, and
a fracture can occur. Here we assume that the fracture initiation pressure is related with the pressure that
results to passive failure of the cavity.
Assuming a linear Mohr-Coloumb criterion, the
plastic stresses must fulfill the criterion:
Figure 7. Definition sketch, for the situation with active and
passive soil around the cavity, see also text.
Where the prime indicates the effective stress.

The material parameters N and S are different for
active and passive yield and can be written as:
and
(The publication of Wang and Dusseault discriminates

between peak values and residual values, here we use
only one value.)
Here is the friction angle and c0 the cohesion of
the soil material. Active yield will occur when:
Table 1.
Input parameters used in calculation.
0 original stress in sand

friction angle sand
c0 cohesion
Rw (radius of tube, see Figure 7)
b (radius with constant press)
100
35
0
0.035
5
kPa
degr.
kPa
m
m
our interest is the pressure at which the passive zone

started.
During unloading, when the active zone is formed
Pr , Based on the equations presented by Wang and
Dusseault, the pressure in the cavity, can be expressed
as a function of the thickness of the plastic zone Ra
and other parameters (see also Figure 7):
Where:
Where pa is the pressure at the boundary of the opening
in the active state (assuming a perfect plastering on this
boundary) and 0 the initial effective stress around the
opening.
When the effective pressure around the TAM
remains higher than the criterion mentioned in Eq. (7),
there will be no plastic deformation and the limit pressure for passive plastic deformation will remain the
same as if there was no unloading on the soil.
In case the criterion of Eq. (7) is fulfilled, there will
be an active plastic zone. Increasing the pressure in the
cavity afterwards will lead to, the situation sketched
in Figure 7. There will be an active plastic zone and
within that a passive plastic zone. When the pressure
is increased further the active zone will disappear, but
When after active yield the pressure is increased the

first soil will come in the passive plastic state when:
Without the active failure this relation would read:
Table 1 presents the input parameters used for calculations using the formulas presented above. Results
265
Consequence may be that the compensation grouting is sand is influenced not only by the grout, but also
to a large extend by the stress distribution around the
bore hole where the TAM is installed. The installation
procedure will influence the shape of the fracture and
the pressure needed to create a fracture.
7
CONCLUSIONS
From our study we came to the following conclusions:

1 Fractures can occur in a field situation where for
comparable soil and using the same grout only cavity expansion is measured in a model test. This can
be caused by the heterogeneity in the field and by
the way the TAMs are installed in the field.
2 Unloading of the soil before it is loaded by injection
of grout will lead to a reduction of the injection
pressure.
3 A reduced injection pressure will lead to a reduced
cake thickness because both bleeding and leak-off
will be reduced.
4 In further experimental research it will be necessary to investigate the influence of the installation
procedure and in-homogeneity of the soil.
Figure 8. Pressure that is necessary to create passive plastic deformation after that first active yield is imposed as a
function of the original active plastic zone.
are shown in Figure 8. Without active plasticity the

pressure where passive plasticity starts is 157 kPa.
From these results it is clear that plastic deformation in
the active state can considerably reduce the borehole
pressure at which passive plastic deformation occurs.
The pressure needed to achieve passive plastic deformation decreases as the active plastic radius created
by the low pressure before the pressure increase is
larger. This presents quantitative information for the
statement already mentioned by Wang & Dusseault:
Our study suggests that initial active formation damage reduces the pressure required to initiate such a
fracture.
6
DISCUSSION
ACKNOWLEDGEMENT
This work was supported by Delft Cluster. The authors
wish to thank The Hubertus Tunnel Combination for
the information presented on the first cross-connection
of this tunnel.
REFERENCES
In Section 4 it was shown qualitatively that only

with a limited filter cake between the grout and the
subsoil a fracture can occur in homogeneous soil.
In in-homogeneous soil the formation of a fracture is
easier, because also the initial deformation will not
be symmetric. For this reason there was a fracture
between the tail void grout and the sand at the Hubertus
Tunnel.
The calculation model by Wang and Dusseault, as
described in Section 5, shows that the pressure to create a fracture is considerably lower in case of a cavity
that has been subjected to plastic unloading, compared
to a cavity that is still in an undisturbed state. This last
phenomenon is the most reasonable explanation that
the injection pressure in the field is lower than the
pressure measured in the laboratory. A lower injection
pressure also means that the filter cake due to plastering and leak-off is thinner and thus that the grout
mixture is less critical. This may be an explanation that
fractures are created in the field for conditions that
does not lead to fractures in the laboratory. Further
experimental work has to prove and quantify this idea.
Andersen, K.H., Rawlings, C.G., Lunne, T.A. & Trond, H.

1994. Estimation of hydraulic fracture pressure in clay.
Journal of Canadian Geotechnical 31: 817828.
Au, S.K.A. 2001. Fundamental study of compensation grouting in clay. PhD thesis. University of Cambridge.
Bezuijen, A., Sanders, M.P.M., Hamer, D. & Tol, A.F. van
2007. Laboratory tests on compensation grouting, the
influence of grout bleeding. Proc. World Tunnel Congress,
Prague.
a bored tunnel. Proc ICOF. Dundee: Thomas Telford.
Bezuijen, A. & Tol, A.F. van 2007. Compensation grouting in
sand, fractures and compaction. Proc. XIV European Conference on Soil Mechanics & Geotechnical Engineering,
Madrid.
Chang, H. 2004. Hydraulic fractures in particulate materials,
Phd. Thesis, Georgia Institute of Technology, November.
Chin, C.Y. & Bolton, M.D. 1999. Factors influencing
hydrofracture in clay. Proc. 13th ASCE Engineering
Mechanics Conference, Baltimore.
Chiriotti, E., Avgnina N. & Grasso P. 2005. Compensation
grouting for TBM tunneling beneath shallow cover. Proc.
5th Int. Symposium Geotechnical aspects of underground
construction in soft ground, Amsterdam 2005.
266
Christiaens M., Hemerijckx E. & Vereerstraeten J.C. 2005.

Tunnelling under the city centre of Antwerp. A new
underground Railway link for the HSL Paris-BrusselsAmsteram. Proc. 5th Int. Symposium Geotechnical
aspects of underground construction in soft ground,
Amsterdam 2005.
Gafar, K. & Soga, K. 2006. Fundamental investigation of
soil-grout interaction in sandy soils. Report. University
of Cambridge.
Gafar, K., Soga, K., Bezuijen A., Sanders M.P.M. & Tol A.F.
van. 2008. Fracturing of sand in compensation grouting. Proc. 6st Int. Symposium on Geotch. Aspects of
Underground Construction in Soft Ground, Shanghai.
Grotenhuis, R. 2004. Fracture Grouting in Theory, Modelling
of fracture grouting in sand. MSc thesis, Delft University
of Technology.
Jaworski, G.W., Seed, H.B. & Duncan, J.M.. 1981. Laboratory
study of hydraulic fracturing. Journal of the Geotechnical
Engineering Division 107(6): 713732.
Kleinlugtenbelt, R. 2005. Compensation grouting, laboratory

tests in sand. MSc thesis. Delft University of Technology.
Kleinlugtenbelt, R. 2006. Private Communications.
Kleinlugtenbelt, R., Bezuijen, A. & Tol A.F. van. 2006.
Model tests on compensation grouting. Proc.WorldTunnel
Congress, Seoul.
Mair, R.J. & Hight, D.W. 1994. Compensation grouting.World
Tunnelling November: 361367.
Mori, A. & Tamura, M. 1987. Hydro-fracturing pressure
of cohesive soil. Journal of the soil and foundations,
Japanese Society of Soil Mechanics & Foundation Engineering 27(1): 1422.
Sanders, M.P.M. 2007. Hydraulic fracture grouting, laboratory tests in sand. MSc thesis. Delft University of
Technology.
Wang, Y. & Dusseault, M.B. 1994. Stresses around a circular
opening in an elastoplastic porous Medium Subjected to
repeated hydraulic loading. Int. J. Rock Mech. Min. Sci. &
Geomech abstr 31(6): 597616.
267
Research of non-motor vehicle-rail transit-tube interchanging

transport system pattern
A.Z.G. Deng & Q.H. Zhang
ABSTRACT: In the light of the relationship between land use and rail transit, it considers that the development
of the public transport is one of the effective ways to solve the problems in traffic. Urban rail transit that can lead
the development of the cities should be the backbone in the public traffic. This paper presents the research about
non-motor vehicle-rail transit interchanging transport system patterns. The interchange pattern is proved to be
feasible by investigation and calculation research. The purpose is to make public transportation more effective,
and elevate the leading role of rail transit. The paper also indicates the effective ways to solve the problems in
traffic in order to exalt the proportion of the urban rail transit.
GENERAL INTRODUCTION
Nowadays, there are many researches about the interchange of rail transit and public transportation (such
as airport, railway, subway) along with the increasing
urban rail transit lines in Beijing and Shanghai etc.
However, these researches hardly mention the interchange of non-motor vehicle traffic to rail transit.
Through the investigation on general distribution station (refers to the station around the inhabited area,
public transportation, non-motor vehicle and walking system primarily), this paper presents the research
about non-motor vehicle-rail transit interchanging
transport system patterns. The purpose is to facilitate
the public transportation, and elevate the leading role
of rail transit. With the development of rail transit network, it is very important to investigate the model of
the affected factor in station and the land use around
to provide a comfortable, safe and convenient environment in transit interchanging. The paper proposes the
effective ways to solve the problems in traffic, in order
to exalt the proportion of the urban rail transit.
2 THEORY OF THE LAND USE AROUND
THE RAIL TRANSPORT STATIONS
The fundamental characteristic of the railway station
can be summarized as the following: forming the
buildings around the railway station, which generate
the core-axis pattern, developing the traffic organization and function layout with the center of railway
station. Land is important to a city, and reasonable use

of land will do benefit to the economy, society as well
as the ecosystem. The aim of land use shows complex, multidimensional and integrative in space form,
which request close cooperation between infrastructure development and civil development to optimize
land use. In the range of railway station, it should be
emphasized of the relation and merge between different city functions, which can make people enjoy
the use of railway and embody the theory of people
oriented.
In developed countries, some new urban planners
break through the limit of the traditional urban planning theory under the citys sustainable development
background, and proposed Transit-Oriented Development Model, i.e. TOD Model. It is expected the use
of public transport to lead the development of a city,
and back to the model of using bicycle and walking.
The main content of TOD is: Using public transportation station as a center, walking distance as a radius.
In this area, high density of the land use is emphasized and public transportation facilities are arranged
around station. By walking, bicycle and public bus etc,
high effective interchange to replace the leading role
of car in city.
Because of the diversity between China and west
developed countries on population density and city
development stage, there are some differences of using
TOD theory, explained as follows:
1. Different background: After the World War II, the
cars become popular in cities of US, the energy
269
waste becomes serious, suburbanization and urban

blight of the city downtown. The TOD concept was
right proposed under this background. But cities
in China are still in highly development. The main
problem is how to develop public transportation,
to solve the traffic jam carried up by high density of citys population and deterioration of the
environment.
2. Population density: In US, low population density, and the main transportation home-based-trips
are cars used. But in China, population density
is higher. Even the residential area in suburb,
the population density is still higher than foreign
country.
3. Willingness to use public transportation: Because
of the distributed suburb, high percentage of people owing cars and more highways, the public
transportation has less attraction to publication in
western countries. But in China, it is because of the
large population, short of effective facilities, and
undeveloped construction technique and management technique that there always exits the traffic
jam, lower speed of transportation and lack of comfortable conditions. There will be a vicious circle
along with the more cars using.
4. The difference of influence radius of station: In
foreign TOD model, there are a 600 meter influence radius with the center of the station. And
the interchange with rail transit is most based on
walk system. But in China, the density of rail transit network is lower, instead of higher population
density, and the influence radius of railway station
is larger, people use interchange mostly by public
bus and bicycle etc with rail transit, which makes
the interchange pattern of the railway station and
surrounding area different.
It can be seen from the operation of foreign big
cities transportation that the percentage of railway
passenger is 45% to 60% to the total one. But in China,
the percentage of bus makes over 75%, and that of railway is much lower. It is because that the development
of rail transit transportation is on the beginning, and
hasnt formed a transport net. The density of station
and network is over 1500 km in London, New York,
Paris and Tokyo, and there is a high railway density in
down town also. The density of railway and stations
in town is shown in Table 1. It will generally take 5
to 10 minutes for people walking from home to the
railway station. Take Beijing for example, the density
of subway in down town is only 0.32 km/km2 , which is
one tenth of that compared to developed countries. It
is pointed out in literature that the downtown district,
middle district and outskirt district of a big city can
be taken as 1.6, 0.8 and 0.4 km/km2 based on Chinas
situation and it is found acceptable. It is still a gap
of the density of the railway system with developed
countries.
Table 1. The density of station and network in London, New

York, Paris and Tokyo.
Density of network
km/km2
Density of station
unit/km2
London
New York
Paris
Tokyo
3.49
3.47
3.76
2.6
2.52
3.92
4.58
1.59
In conclusion, due to low network density in china,

the application of TOD model in Chinas cities will
be modified as: it should be based on the station as a
center, rail transit, bus and non-motor vehicles as an
interchange for development. The main task is not to
increase the residential density but how to combine
land use and public transport together, which can provide a comfort, convenient and attractive public transport system. By doing so can we make use of the network effectively as the backbone in citys development.
3
RAIL TRANSPORT SYSTEM PATTERNS IN

DIFFERENT CHARACTERISTIC STATION
To different railway transportation types of stations

(It is classified into large scale interchange station,
general interchange station and general distribution
station) area, accordingly there are different major
functions (seen in Table 2). Generally, for station
in downtown area (large-scale interchange station or
general interchange station), the commercial business
development area consist up the majority of this district, and the residential development area takes up
the minority of this district. For station in outskirt or
newly developed area (general interchange station or
general station), the most suitable development in this
area should be residential development.
In planning the station and surrounding area, the
main concern point should be the land use around
the station. For example, in city CBD area, the station
should be built as the commercial center of, especially
in interchange station of multi-trip railway lines. The
underground space exploitation should also be considered to develop a district, which is incorporating
transportation and business or other functions to form
an underground urban complex.
For interchange station in outskirt of the city or in
development zone, it should be considered together
with other transport method e.g. private cars, buses,
non-engine cars.
For stations at suburbs of the city, the majority
method of interchange is using Cars Peripheral Park
and Ride (P&R). The P&R refers to the interchanging
facilities for people in outskirt parking cars around the
railway station or public transportations end stop. By
using public transportation system to down town, the
270
Table 2.
Classification in different rail transit station.

Classification in different
rail transit station
Rail transit
station
Large-scale interchange
station
Land use around station
Different interchange forms
Town center commercial

circle and CBD area
City suburb
City distribution center
Within rail transits (3 or

over 3 lines)
Rail transit with air plane
Rail transit with railway
Rail transit with Road passenger
transportation
Rail transit with public transport
commercial area and

residential district
General interchange
station
General distribution
station
City sub-center
Development zone
Residential district
use of private cars in down town area is decreased.

The outskirt interchange facilities are the link of cars
and public transportation system. It major function is
to provide the effective, safe, convenient, comfortable
linkage between outskirt cars system and downtown
public transportation system, so as to increase the
attraction of the public transportation system and
decrease the use of cars in down town area. By doing
so, the transportation pressure in down town area could
be diminished and realize the sustainable development
of the citys transportation system. This method is suitable for outskirt of the city. The railway station must
provide the sufficient parking facilities, parking lots
to meet the interchange need. The parking facilities
must be near to the station, and have connection path
to the station. The enough roads should be planned and
constructed.
Regarding in urban general distribution station, preceded by text analysis, the main consideration is the
walk system, the non-motor vehicle transportation
system and the bus, the taxi convenient interchanges.
The walking system is the most primary connection with the rail transit, the content mainly includes
sidewalks system in station, the facility and the passengers separates from vehicle facility plan design,
the guidance informational sign design, walking routes
organization design and so on. Therefore the walking
system should be given priority in rail transit design.
Considered transportation station around the land use
intensity is high, passengers activity is also frequent,
and in order to increase commercial stores suitable
sidewalks should be provided. Traffic island and crosswalk must be considerable and designed in this area.
The walk system in the station which connect platform meets the convenient needs in the station, but
also achieve the evacuated request, simultaneously
also must have distinct guidance symbol.
Within rail transits (2 lines)

Peripheral Park and Ride (P&R)
Rail transit with non-motor vehicle
Rail transit with walk system
The bicycle is one of effective ways connected with

rail transit also. Enough bicycle parks should be provided in joints of station design, appropriative parking
zone be setting which is connected with underground
tunnel, accommodation road should be constructed in
order to lower the effects from non-motor vehicle in
traffic.
The interchange between bus transportation and
rail transit mainly includes the roadside parking pattern, parallel pattern, vertical pattern and centralized
pattern. Bus accommodation road can be presented
and the bus stop should be built near subway station
entrance-exit which can provide comfortable, safe and
convenient environment in transit interchanging.
RESEARCH OF NON-MOTOR
VEHICLE-RAIL TRANSIT-TUBE
INTERCHANGING TRANSPORT
SYSTEM PATTERN
Bicycle is one of the effective manners in public traffic.

Especially in china at present, inferior service in public traffic, undeveloped mechanization. Bicycle shows
cheap, convenient and unpolluted characteristics, it
takes important status in our country public traffic. The
interchanging content mainly includes bicycle park
system in station, the facility of the passengers separates from vehicle facility planned and suitable routes
organization designed near the station.
Bicycle-rail transit-tube transferring transport system patterns design should follow principles:
1. The interchange pattern is suitable in the city suburb
or in residential district around rail transit station
(General in distribution station), and it can be not
applied in city center.
271
2. In order to avoiding occupying of the space,

decrease the effects to the transport, it should provide enough quantity if the bicycle special parking
spots.
3. Appropriative bicycle parking lots on the ground
should be provided near the station which passenger
flow is large, and passengers may take interchange
through tunnel; dispersed bicycle parking zone may
be designed around the station which has few passenger flow, yet is not so closed to the station
entrance-exit in order to avoid effecting passengers.
4. In park essential facilities must be supplied and
should arrange the specialist to manage with inexpensive charge.
5. To display bicycle superiority in short-distance
home-based-trips, and limits its proportion in
long-distance home-based-trips. Reasonable bicycle routes and accommodation road system can
reduce the bicycle effects in traffic and provides
comfortable, safe and convenient environment in
transit interchanging.
In general distribution station, which interchange
with rail transit almost are public traffic (bus) and
non-motor vehicle (such as automobile, bicycle and
walk system). At present in many stations usual show:
bicycles are laid on the street or some temporary openair non-motor vehicle parks spot all around, it has
occupied the limited station space. Simultaneously
the bicycle around the station may on the passenger
way, hinder the pedestrian traffic, and potential security risk exists also. Although many stations peripheral
non-motor vehicle parks have take specialists, accommodation tunnel which may connect rail transit and
bicycles are not been designed. It enlarges interchanging distance, takes inconvenient and lowers the
interchanging efficiency.
Based on the interchange principle between nonmotor vehicle and rail transit introduced above, it is
proposed in this paper the Bicycle-rail transit-tube
transferring transport system patterns concept by using
a general underground subway station as an example.
For a general station, the length is about 200 m,
the width is about 20 m (3035 m if including exit).
Four exits are designed on both sides of the station,
and are constructed in conjunction with the ventilation
shaft etc affiliates as shown in Figure 1. The station is
situated under the cross, and four exits are situated at
four corners of the road. In construction of the station,
the main structure will be constructed first, and then
the exit. It can be seen from the plane layout that the
two exits in the same direction would be constructed
separately, and no connection between the two exits.
The shadow area in the middle has no use. If it can
be constructed at the same time with the exits of the
station, the two exits and the middle part could form
the underground parking bicycle etc. By doing so, the
land around the station can be used intensively, and
Figure 1. Bicycle-rail transit-tube interchanging transport

system patterns in General distribution station on the 1st flood
underground.
Figure 2. Sketch map of Bicycle parks underground.
decrease the station ground area occupation area and

passenger flow on the ground. The phenomenon of
scramble between bicycle and car will be decreased,
and the interchange efficiency of non-motor vehicle
and railway will be highly increased.
The underground non-motor vehicle is shown as
shadow area in fig. 1, the four corners of the road
could be set as bicycles parking exit (shown in Fig. 2)
which makes the passengers from both directions can
park their bicycles effectively and interchange with rail
transit (interchanging may use the mutual wall between
station hall and bicycle parking area or use tunnel). For
consideration of parking area of bicycle, this shadow
area is about one third of the station main area (about
60 m) with the width of 20 m, area of 1200 m2 (total
area 2400 m2 ). The major design reference can be
planed by Bicycle Parking Design Standard.
The principle factors considered for the scale of a
normal park for bicycles include the number of bicycles reaching to station, the area occupied by one
bicycle, and the piece of the bicycle park. According to the piece, the scale of the bicycle park can be
calculated as following:
272
Table 3. The main design index in bicycle park.

Parking width
Parking type
Diagonal
30
45
60
Vertical
single
(m)
double
(m)
Space
between
bicycles
(m)
Aisle width
single
(m)
double
(m)
1.00
1.40
1.70
1.60
2.26
2.77
0.50
0.50
0.50
1.20
1.20
1.50
2.0
2.0
2.0
2.00
3.20
0.60
1.50
2.0
Unit area
Parking type
Diagonal
Vertical
30
45
60
Single
one-side
Single
two-sides
Double
one-side
Double
two-sides
m2
m2
m2
m2
2.20
1.84
1.85
2.00
1.70
1.73
2.00
1.65
1.67
1.80
1.51
1.55
2.10
1.98
1.86
1.74
Figure 3. Sketch map of Bicycle-rail transit-tube interchanging transport system patterns (station on one side of
road).
where: Sbic denotes the needed scale for bicycles in

the station (m2 ); Vbic is the number of bicycles during rush hour piece/hour); bic is the velocity of the
bicycle park; sbic is the averaged area occupied by a
bicycle (m2 ).
The velocity can be set as 1 due to little change of the
number of bicycles because the users of the bicycles in
the station are almost only to work or back from work
by bicycles. The area occupied by a bicycle, obtained
fromTable 3, is 1.7 m2 .Through calculating, the capacity of the underground bicycle park, excluding the
public area, is 1200 units.
Through the sampling survey of numbers of bicycles near the available stations in Shanghai rail transit
carried out by the author, the following point are
given: the number of persons using bicycles for interchanging transportation, including the persons who
arbitrarily lay their bicycles at the edges of road or
near some shops, is about 800 to 1000. According to
the above results, it can be seen that the capacity of the
underground park area in a normal station can meet
the requirement of paring bicycles for transportation
interchanging.
For the station of main body lay on the side of road,
the underground park area for bicycles can be set up
as shown in Figure 3. For this type of parking area, the
ceiling can be mounted by transparent materials for
accessing of sunlight. The vegetable processing on the
top of the park will improve the inner environment in
an extent through the hole between the parking area
and station hall.
Taking the variety of the plane layout of the railway

transportation stations into accounting, the underground parking area for bicycles combined railway
station, proposed in this paper, is a conceptual pattern.
The exact arrangement will be adjusted according to
the actual plane layout of stations. The principle idea
of the model is to efficiently utilize the area near the
entrances and exits, realize the interchanging between
bicycles and railway transportation without extra cost,
and increase the efficiency of interchanging.
For the underground parking area of bicycles in a
station, the construction is feasible in techniques. In
general, the stations which need this type of parking
area are in the sub-center, outskirt or near the satellite
city. The main buildings near the stations are residential districts also, the construction of the station will
almost not affected by the environment around. The
parking area can be constructed with the main structure
of the station by open excavation method in foundation. Also, its feasible to firstly construct the main
structure of the station and then construct the parking area combined the entrances and exits by open
excavation method or top-down method.
CONCLUSIONS
Through the analysis above, it can be concluded that

the underground bicycle parking area for transit interchanging is feasible in reality. The advantages as
below:
1. The users of bicycles can interchange rail transit
through the entrance between the under ground
parking area and station conveniently, which realize the interchange without extra cost, increasing
the efficiency in a great extent, providing the comfortable environment in transit interchanging at
same time and expressing the conception of people
oriented.
273
2. The underground interchange between bicycle and

railway transit decreases the occupied area of the
station, facilitate the travelers using other types of
transportation which transfigure the environment
around the station.
3. This pattern of transit interchanging can decrease
the flow of bicycles on the ground, increases the
efficiency of motor passing, eliminates the possibility of traffic accident and realizes the Pedestrian System Separated from Vehicle System by
passenger interchanging through tunnels.
The cost of construction for the underground bicycles parking area is higher than normal station.
Nevertheless, synthetically considering the Traffic
Diversion on the ground, the effect on the passenger
flow by other patterns of transportation, and the environment near the station, the society and economic
benefit resulting from what it brings will exceed far
from what it costs.
REFERENCES
Ge, L. 2005. A Study on Planning & Design Methods for
Urban Passenger Transit Hubs. Doctoral Dissertations In
Southeast University. Nanjing.
Gu, B.N. & Rao, X.P. 2001. On the Reduction of Subway
Station Scale. Urban mass transit. 2001(3):1417.
Huang, J.Z. 2006. Models for Suburbanization Land use and
Corresponding Transportation Development in Metropolis. Beijing: China Architecture & Building Press.
Yang, D.Y. & Han, H. 2000. A Study on Metropolitan Rail
Transit and Transportation Structure. Urban mass transit.
2000(4):1015.
Zheng, M.Y. 2006. Urban Develop in Light-RailAge. Beijing:
China Railway Publishing House.
274
Shotcrete excavations for the Munich subway Comparison of

different methods of face support in settlement sensitive areas
J. Fillibeck & N. Vogt
Zentrum Geotechnik, Technische Universitt Mnchen, Mnchen, Germany
ABSTRACT: For the construction of shallow tunnels in settlement-sensitive urban areas it is very important
to reduce the settlements and to increase the stability of the tunnel face during the excavation. In the case of
shotcrete excavation, the use of different methods of face support has become more and more common. These
methods are: ground freezing, pipe roofs, jet grouting and injection support. The paper shows the experience
made to due the installation of the above mentioned face supports, especially since the specific focus is related to
the arising settlements. If only small deformations are allowed to occur, as the examples show, deformations that
have to be considered during the construction process as well as those to establish the bearing load; they could
be significant depending on the process. Suggestions have been made as to ways in which the deformations can
be reduced by making additional measurements.
INTRODUCTION
For the construction of safe shallow tunnels in

settlement-sensitive urban areas, it is very important to
reduce the settlements and to increase the stability of
the tunnel face during the excavation. The use of different methods of face support in the case of shotcrete
excavations is becoming increasingly common. These
methods are ground freezing, pipe roofs, jet grouting
and injection support.
The report presents the experience gained in the
installation of the abovementioned working face supports, with particular focus on the induced settlements. Four different projects of Munichs subway
are described. After a short project description the
results of the measurements are illustrated (geodetic and borehole measurements) and evaluated. With
this background, the different methods of face support are compared and the different advantages and
disadvantages are discussed. Finally, special attention
is focused on the installation process. Proposals are
made for the reduction of settlements in future tunnel
projects.
laminated and have, depending on the deposit conditions and their age, a differing amount of sand and fine
grain. The average permeability amounts to approx.
k = 5 103 m/s. Tertiary layers lie below the quaternary gravels. They consist of changing layers of fineto medium-grained sands with high density and clays
or silts in stiff to firm consistency. The thickness of the
layers can change excessively within a small distance.
The average permeability of the sand amounts approximately from k = 1 104 to 1 105 m/s, the tertiary
clay and silt can for all practical purposes be assumed
impermeable.
The quaternary gravels possess a mostly free
phreatic water level, which can reach ground level.
There are still confined aquifers within the sand layers
with fine-grained cover. The pressure of the groundwater approximately corresponds to that of the free
phreatic surface in the quaternary gravels.
3
3.1
GEOLOGICAL AND HYDROGEOLOGICAL

CONDITIONS
In the Munich subsoil the quaternary gravels follow

under fillings of small thickness. The quaternary gravels can reach a thickness of more than 20 m. They
predominantly consist of medium density layers, are
HEADING WITH GROUNDFREEZING

UNDER THE CITY HALL OF MUNICH
Construction process
The extention of the station Marienplatz of the subway

lines U3/ U6 under the Munich City Hall was built by
the company Fa. Max Bgl GmbH & Co KG. The
project was finished in 2006.
Parallel to the two existing platforms, two directly
joining tunnels were built in shotcrete method under
atmospheric conditions with a vertical distance of
275
border freezing area
intermittent
operating
permanent
operating
30
20
10
0
-10
core freezing area
-20
permanent
operating
-30
-40
0
20
intermittent
operating
40
60
80
100
days
Figure 1. Cross section with City Hall, tunnels and geologic

situation.
Figure 2. Operating Control of the artificial ground

freezing.
In order to reduce frost heave to a minimum the

following measurements had to be taken:
about 10 m to the city hall. In order to avoid damage to the landmark city hall, the deformations had
to be strictly limited. The construction company Fa.
Bgl, planned freezing arches in the context of an alternate bid, in order to support the crown and to keep the
retaining water away from the tunnel face. The freezing arches were provided for through pilot galleries
above the crown (Figure 1).
The tunnels are embedded in the tertiary layers
(figure 1). The water bearing sand layers had to be
dewatered with the help of filter wells.
3.2
Measurements to reduce frost heave
For the successful realization of the specific proposal

it was crucial to reduce frost heave in such a way that
no damage occur to the city hall. Frost heave can be
essentially attributed to two reasons:
homogenous frost heave (hvol ) because of a 9%
increase in volume caused by the changeover from
water to ice.
growing of ice lenses with corresponding frost
heave (hicel. ) because of the tendency of the soil
to draw water near the interface of the frozen to the
unfrozen soil (zero-degree-front). This frost heave
increases with time.
The frost heave tests, which were performed in the
laboratory of the Centre for Geotechnics at the Technical University of Munich (Zentrum Geotechnik, TU
Mnchen), showed, that in the tertiary fine-grained
soils frost heave hicel. still occurs at load-levels
of more than 400 kN/m2 if water can be drawn at
the interface to the permeable sand layers. Therefore
the alternating layers of permeable sands and frostsensitive clays present a critical risk source when frost
heave is considered.
measuring and controlling the temperature in the

soil with the help of 5 measuring cross sections
per tunnel. Every cross section includes 18 thermo
couples.
reducing the operation time of the frozen arches by
dividing the tunnels into 3 different sections: north,
middle and south.
further partitioning within the freezing sections by
the installation of groups of freezing tubes with
separate control.
In figure 2 the temperature development of a
section is shown schematically. Twenty days after starting the freezing process in the core of the freezing body
the freezing of the border area started. After the frozen
body reached approx. 22 C in the core area and 0 C
in the border area, the freezing tubes were operated
intermittently, an average of 8 to 24 hours in the core
area and 12 to 24 hours in the border area. Due to intermittent handling, the zero-degree-front does not move
outside (enlargement of the frozen body), but stays
in a narrow zone, which again and again gets frozen
and defrosted. Thus frost heave reduces significantly.
The operation of one section could be stopped after
approx. 90 days. The adjacent defrosting process took
about three months.
3.3 Measuring of the settlements
The deformations which occurred during the construction process were measured by a geodetic precise
levelling system on the surface and a closed water
levelling system in the 2nd basement of the city hall.
The closed water levelling system consisted of 10
measuring points with a resolution of 1/10 mm. The
measuring results could be checked online at all times.
Figure 3 shows the location of the measuring points
S03, S06 and S09 of the closed water levelling system
as well as the development of the settlement and heave
276
Figure 3. Vertical displacements of the measuring points

S03, S06 and S09.
depending on the time. The three measurement points

were situated in the sections north, middle and south.
The settlements at the beginning of the freezing process result from groundwater drawdown. At the onset
of freezing the expected frost heave started. It reached
a maximum value of 3 to 5 mm. The settlements due to
the tunnelling process occurred after the heading had
passed the measuring points and they still continued
after the freezing process was stopped. The settlements
slowed down continuously and stopped 3 months later
with a maximum settlement of about 10 to 12 mm.
Figure 3 clearly shows the temporary displacement of the settlements according to the heading.
The drive reached the measuring points in descending
order, resulting in them reaching the maximum heave
successively.
The measured deformations were approximately the
same as the calculated ones. Thereby half of the settlements could be attributed to dewatering measures,
which lead to large area settlements and correspondingly low differential settlements. Furthermore, no
settlement damages were determined at the city hall,
so it can be assumed that the heading was very successful. It was essential that for the success of the project,
larger frost heave by ice lenses could be avoided by
applying the above mentioned measures. Frost lenses
would otherwise have led to a softening of larger soil
areas and therefore to larger settlements and settlement
differences.
4
4.1
JET GROUTING COVER FOR A LARGE

CROSS SECTION FOR U3 NORTH LOT 1
Figure 4. Heading W3 and W4 with jet grouting cover.
Figure 5. Deformations in cross section MQ 8 after construction of the jet grouting cover.
compressed air support and with several methods of

crown support which will be introduced hereafter.
The headings W3 and W4 with a cross section area
of up to 200 m2 began from a starting shaft with a top
heading. First watertight pits with thin slurry walls
were produced (figure 4) in order to lower the groundwater table. The safety of the excavation face was
increased by 13 jet grouting covers (total length of
about 15.5 m each, overlap 4.3 m) as well as further jet
grouting piles in the face of the crown.
The quaternary gravels were cut with suspension
(simplex method) at a pressure of up to 400 bar at the
cone. At anytime during the making of a jet grouting
pipe a controlled outflow of the suspension is required,
assuring that the pressure does not lift the soil. The
top heading followed after the installation of the jet
grouting took place. The heading of the bench and
invert began on finishing all top headings.
The consortium Ed. Zblin AG/Max Bgl GmbH &

Co KG carried out the construction of the subway lot
U3 North 1 in the north of Munich. The works were
completed in 2006. The shotcrete headings with a total
length of around 1950 m were driven with and without
4.2 Heave during the installation of the

jet grouting cover
Figure 5 shows the deformation in the cross section at
a distance of 50 m away from the starting shaft directly
after the installation of the jet grouting cover.
277
Figure 7. Comparison of settlements of different shotcrete

tunnels, driven under atmospheric conditions.
4.3 Settlements during the heading

Figure 6. Required suspension backflow during jet
grouting.
The heave above the crown reached in MQ 8 about

140 mm and in total a maximum of 250 mm. At first,
the heave was deemed uncritical because there were
no buildings close to the tunnel, however they result
particularly in soil strains in a narrow band directly
above the interface to the tertiary soils. Because of the
proximity of the thin slurry wall to the jet grouting
cover, the heave led to a crack in the thin slurry wall,
making the wall permeable.
The heave results from the fact, that the outflow of
the suspension in the annular space of the jet grouting
piles, which are faced upwards, can not be controlled in
a suitable way (figure 6). Owing to the lack of backflow
in the layered soil with strongly differing conductivity
the overpressure spread over a greater area, resulting
in a rising of the soil above the jet grouting cover.
The large heave could only be limited by reducing the overpressure through the installation of further
boreholes from the ground level, resulting in high
costs.
With increasing soil cover the heave reduced
because of the increasing load. However, even under
more than 12 m of soil cover and the installation of
the jet grouting cover in tertiary clays, the heave still
amounted to approximately 20 mm.
For further projects, where only very few deformations are allowed during the installation of the jet
grouting cover, sufficient attention should be paid to
the control of the suspension backflow during grouting. This problem could for example be resolved by
improving the technique of the grouting machine or
with the help of a double tube, which is pulled a little
ahead during jet grouting or likewise with the help
of special valves which control the pressure of the
backflow.
If the pressure gets to high during grouting, heave
can be avoided or at least reduced by additional
horizontal or vertical arranged boreholes.
After the installation of the jet grouting cover, the top

heading with temporary shotcrete invert followed step
by step, over the whole heading distance. After this the
heading of the bench and invert followed. The settlements which occurred during the headings amounted
to a maximum of 26 mm in cross section MQ 8 and 30
to 40 mm in the area with larger soil cover.
In order to be able to judge the results, in figure 7
the above mentioned measurements are compared
with those of atmospheric shotcrete headings having
nearly the same soil cover but were driven in partial
face advance without jet grouting cover.
Overall maximum settlements were measured as
almost the same size, which means that the jet grouting
cover does not reduce the settlements, in comparison
to tunnels driven in partial face advance. As the sliding micrometer measurements show, the forces which
were taken from the jet grouting cover, lead to concentrated high stresses in the small bedding area of
the jet grouting cover. This stress concentration leads
to comparatively high compressions and settlements.
On the other hand partial face advance leads to less
stress concentration, however the different delayed
headings lead to multiple load rearrangements and
therefore the surface experiences approximately the
same settlements.
Finally it can be concluded that with the jet grouting
cover the settlements are not reduced in comparison to
those caused by partial face advances. However, the
face stability clearly increases by using a jet grouting
cover.
PIPE SCREEN COVER FOR THE

UNDERPINNING OF A BUILDING IN U3
NORTH LOT 1
The two shotcrete headings of section W1 in the above

described subway Lot U3 North-1 in Munich had a
cross sectional area of A = 41 m2 and were driven in
the tertiary soils under atmospheric conditions with
the help of wells, dewatering the tertiary sand layers.
278
Figure 10. Longitudinal section of the grouting section O2.

Figure 8. Crossing the building Werner-Friedmann-Bogen,
longitudinal section.
Figure 9. Settlements of Werner-Friedmann-Bogen during

tunnelling (cross section).
In this section the underpinning of the WernerFriedmann-Bogen, a building complex with 12 floors,
is of special interest. The foundation pressure of the
3 m wide strip foundation, which lies in the centre of
the building and carries the main loads, amounting to
nearly 300 kN/m2 .
At a vertical distance of approx. 12 m between
the foundation and the crown, a pipe screen cover
was planned as an additional measure of protection
(figure 8), because the tertiary soil cover amounted
only 4 m and full water pressure was acting from the
quaternary to the surface of the tertiary soils. At the
southwest side of the Werner-Friedmann-Bogen an
underground garage follows.
For every pipe screen 38 pipes were installed. The
length of the pipes amounted to 12 m with an overlap
of 4 m. The bore diameter amounted to 146 mm with
a 6 mm annular space.
In figure 9 the settlement trough along the WernerFriedmann-Bogen is shown as dependent on the
development of the heading.
Due to the dewatering of the tertiary sand layers

settlements of 5 mm to 7 mm were recorded.The installation of the pipe screen and the forward directed
settlements of the heading of track 2 increased the
maximum settlements to approx. 10 mm. The largest
settlements resulted from the 2 headings. Finally the
maximum settlements amounted to 25 mm.
As a comparison with measuring of further cross
sections without pipe screen shows, the maximum settlements were measured under the Werner-FriedmannBogen. It is clear that the foundation loads lead to
higher settlements and because of the smaller soil
cover only limited arching develops. Furthermore,
installation also cause settlements. However, it is crucial to settlements, that the pipe screens as well as
the surrounding soil layers experience some deformation, before the system can carry the expected load
in both the longitudinal and lateral directions. That is
why the predominant settlements respectively occur
shortly before and directly during the heading.
It can therefore be concluded, that the pipe screen
primarily increases the safety of the tunnel face. For the
installation and formation of the bearing effects, deformations are however necessary, which lead in this case
to settlements of 25 mm. Pipe screens are only applicable for the reduction of settlements, if substantially
higher settlements are expected without them.
HEADING WITH COMPRESSED AIR

SUPPORT AND GROUTING IN U3 NORTH
LOT 1
The geological conditions in the section O2 of the subway lot U3 north-1 are shown in figure 10. In this
section a shotcrete heading with compressed air support was provided for. If the thickness of the tertiary
soils above the crown reached less than 1.5 m, the overlaying quaternary gravels were grouted. In the grouting
section 1 with a length of approx. 40 m the gravels were
grouted from the surface. In section 2 the surface was
not accessible. The gravel was therefore grouted from
279
Considering that the measured compressed air consumption was almost just as small as the calculated
value, the very extensive grouting measure (21500 m
grouting boreholes with more than 43000 grouting
sleeves have been installed) can be determined as very
successful.
7
Figure 11. Surface settlements due to shotcrete tunnelling

using compressed air in Munich with dependence on the pillar
ratio A/D.
the tunnel. At the end of the heading the grouting was

done again from the surface.
The aim of the grouting was to reduce the permeability of the gravel to k 5 105 m/s. This was
controlled by permeability tests in the bore hole.
After the grouting activity the heading followed in
shotcrete method with compressed air support with a
maximum overpressure of 0.7 bar. Settlements were
measured between 3 mm and 11 mm (without considerating of settlements due to water drainage). In order
to assess this result, the results of settlement measurements of headings in Munich, with compressed air
support and without grouting (shield tunnelling and
shotcrete method) depending on the pillar ratio A/D
are compared with the abovementioned result of the
U3N1 measurement in figure 11.
It can be seen that the settlements measured if
grouting was applied do not differ from those without
grouting. It appears that the grouting did not reduce
the settlements.
Overall, the results confirm, that shotcrete headings
with compressed air support lead only to very small
settlements with small tangential inclinations, which
cause no damage to conventional buildings.
CONCLUSIONS AND FINAL REMARKS
In order to construct shallow tunnels in settlementsensitive urban areas with the shotcrete method, measures have to be taken to increase the face stability and
to reduce the settlements.
Besides the common measures (for example reducing the length of the advance step, etc.) special crown
support measures are often used for this purpose. In
this paper, the experiences of the authors demonstrated
the purpose of using a frozen cover, a pipe screen cover,
a jet grouting cover and a grouting cover. It has been
shown, that the installation of crown supporting measures can have an extensive influence on the appearing
deformations
As the settlements caused by shotcrete headings
with conventional cross sections (approx. 40 m2 ) in the
Munich underground are comparatively small (smaller
than 20 mm to 25 mm for atmospheric headings and
smaller than 10 to 20 mm for headings with compressed air support), the crown supporting measures
do not have, as the examples show, decisive advantages
regarding the deformations. However, if the crown
supporting measures are used in a adequate way, a considerably higher safety potential occurs. This has to be
considered if a decision has to be made, as to whether
or not crown supporting measures are necessary in
difficult sections.
280
Fracturing of sand in compensation grouting

K. Gafar & K. Soga
University of Cambridge, Cambridge, UK
A. Bezuijen, M.P.M. Sanders & A.F. van Tol

TU-Delft/GeoDelft, Delft, The Netherlands
ABSTRACT: The phenomenon of fracturing in sand as a result of compensation grouting was studied. Processes
of fracture initiation and propagation were explained and a parametric study was conducted in order to investigate
the factors that cause sand fracturing to occur. Experimental results indicate that fracture initiation requires the
existence of a local inhomogeneity around the injection position. Grout mixture in terms of water-cement ratio
and fines content had major roles in sand fracturing, whereas injection rate had a minor influence under the
tested conditions.
INTRODUCTION
Compensation grouting has been widely in use to control ground settlements during tunneling processes.
Nevertheless, its use is still hindered in many cases
by the uncertainties in the grout mechanical behavior.
As the current grouting practice is highly dependant
on field experience rather than scientific knowledge
of soil-grout interaction behavior, issues such as suitable injection pressure, soil fracturing and bleeding
(amount of water forced out of the grout mixture) still
need further investigation.
In particular, soil fracturing stands out as one of
the major challenges that could affect the results
of a grouting project. Accidentally-created fractures
could cause considerable damage to near-by structures, whereas failing to create fractures when they are
required could result in tunneling-induced settlements
(for example) not being fully compensated for.
While fracturing of cohesive soils was extensively
studied by many researchers (e.g. Jaworski et al. 1981,
Mori & Tamura 1987, Andersen et al. 1994, Chin
& Bolton 1999, Soga et al. 2005 & 2006), there
has been limited work for compensation grouting in
cohesionless materials (e.g. Chang 2004). Fracturing
of sand was studied in relation to the oil industry
(Khodaverdian & McElfresh 2000, Bohloli & de Pater
2006) and horizontal directional drilling (Bezuijen
et al. 2002). Recently, the prospected use of fracture grouting in Amsterdam to tackle settlements
encountered during the construction of the NorthSouth Metro line triggered a thorough research into
the phenomenon of sand fracturing.
This paper follows on from the work reported by

Sanders (2007) and Bezuijen & van Tol (2007). Utilizing the reported optimum grout mixture for soil
fracturing, a series of laboratory scale grout injection
tests was performed in which various factors affecting
fracturing of sand were studied.
2 THEORY & BACKGROUND
Hydraulic fracturing is defined as the condition leading to the creation and propagation of a thin physical
separation in a soil or rock mass due to high fluid
pressures. The fracturing process is mainly characterized by fracture initiation, propagation and orientation.
Understanding the factors controlling these parameters is the first step in understanding and predicting the
grout fracturing behaviour. The factors affecting fracturing phenomenon could be divided into two groups:
factors related to soil, such as soil properties (particle
size, shape and distribution, relative density, cohesion,
friction angle etc), stress state and the magnitude of
confining pressure and factors related to grout itself
and the grouting process, such as grout rheology (components and viscosity), injection rate and injection
pressure.
2.1
Fracture initiation
The two main theories explaining fracture initiation

are tensile failure and shear failure. Jaworski et al.
(1981) suggested that, for a hydraulic fracture to occur,
the effective stress has to become tensile and equal in
281
Figure 1. Realistic particle arrangement around an injection hole and possible deformation modes. Pf is the injection
pressure to cause plastic deformations in soil and in the
effective stress around the injection hole. (Bezuijen & van
Tol 2007).
magnitude to the tensile strength of the soil. This situation is clearly unconceivable in case of cohesionless
soils like sand. The other theory on fracture initiation
proposes a shear failure as the main reason for fracturing in clays. Mori & Tamura (1987) suggested that
shear failure occurs within a short duration under a
high injection rate, with the duration being too short
for the grout to penetrate into micro fissures to create
wedge action and/or enter the soil pores to weaken the
soil strength.
For fracturing in sand, Bezuijen et al. (2007) and
Bezuijen & van Tol (2007) suggested that the local
contact forces between sand grains have to be eliminated. As in case of cohesionless soils there is no
tensile strength between the grains, this means that
the fracturing pressure has to overcome the effective
stress in the direction perpendicular to the fracture.
In reality, sand is never perfectly homogeneous and
therefore, the arrangement of particles around a certain boundary (injection hole, for example) will be in
such a way that some particles are in closer contact
than others, as shown in Figure 1.
When injection is conducted, grout will fill the
space between particles and start to push them apart.
Whether this initiated fracture will propagate further or
the result will be a roughly symmetric cavity expansion
depends mainly on the properties of the injected grout.
This explanation of fracture initiation in sand agrees
with the findings of Thallak (1991), who, based on
the micromechanics of granular media, suggested that
hydraulic fracture initiation requires the local contact
forces between particles to become zero or tensile. As
this requirement is complicated by local force distribution at microscopic scale (which can lead to a wide
variation in the fracturing pressure), it was concluded
that hydraulic fracture initiation depends mainly on the
local microscopic inhomogeneities in the soil.
2.2
Fracture propagation
Starting from the condition shown in Figure 1, a

roughly symmetric cavity expansion will always be the
Figure 2. Leak-off and filter cake formation associated with

fracturing.
result if the injection pressure is high enough and close

to perfect cavity expansion pressure. Factors that dictate how high the injection pressure will be are mainly
the grout materials and rheology, confining pressure
and stress state and soil density. Results reported by
Kleinlugtenbelt (2005), Gafar & Soga (2006), Sanders
(2007) and Bezuijen & van Tol (2007) confirm that
high injection pressures are associated with high grout
viscosities (low w/c ratio or more cement content for
cement-based grouts). Gafar & Soga (2006) reported
that injection pressure in the case of no confinement
was increased by 15 times when a 100 kPa confinement was introduced. Bezuijen & van Tol (2008)
highlighted the influence of the stress state. Unloading of the soil around a cavity as a result of equipment
installation leads to plastic deformations in the soil,
which may result in lower injection pressures.
Bezuijen & van Tol (2007) explained that, for fractures to propagate, grout mix has to contain enough
content of fine bentonite particles and has enough
water to ensure good flowability at the same time.
In this case, pressure application will cause the water
in adjacent sand pores to be replaced with a mixture
of water and finer particles leaking from the grout
mix, as shown in Figure 2. This permeation action is
termed as leak-off. With introduction of fines in the
soil matrix, the leak-off will reduce the permeability of
sand around the injection hole, causing grout bleeding
to be slowed down.
On the sides of the propagating fracture, a filter
cake is formed as a result of the accumulation of larger
cement particles filtered at the sand-grout boundary.
Filter cake formation is crucial for the propagating
fracture to be able to keep itself open and to sustain
forces induced by grout penetration. At the same time,
bleeding is restricted, keeping a good workability of
the grout. A suitably high initial water-cement ratio
(w/c ratio) will ensure that the grout mix will have
282
Figure 3. Schematic diagram of the experimental setup,

showing the dimensions in mm (Bezuijen & van Tol 2007).
sufficient water content to keep a fracture propagating

without the need for very high pressures. Eventually, the pressure at fracture tip will not be enough
to overcome the effective pressure in the direction perpendicular to fracture. Nonetheless, the pressure at the
tip will still be high enough for leak-off and bleeding
to continue. This will cause a filter cake to be formed
at the fracture tip, completely blocking further grout
propagation.
3
Figure 4. Simplified model of Tube Manchette (TAM)

used for injection. Actual space between the two rings is
40 mm during injection (Bezuijen et al. 2007).
EXPERIMENTAL SETUP
Grout injection tests were conducted in a cylindrical

steel container of a 900 mm diameter and changeable
height. Two sample heights were used: 840 mm (for
Tests 1 and 2) and 600 mm (for the rest of the tests). Figure 3 shows a schematic diagram of the experimental
setup. The injection tube position was fixed at 360 mm
above the bottom of the container. A PVC plate rests
on the top of the saturated sand sample, tightly sealing
it off from an upper water chamber. Confinement is
applied by means of pressurizing this water chamber
using compressed air. Air pressure is applied through a
glass cylinder that also shows the change in water level
resulting from soil heave. This change is continuously
measured during the test by means of a differential
pressure gauge. Two tubes connect the top (through
the water chamber) and the bottom of the soil sample
to another graduated glass cylinder which rests on the
top of the setup, providing a double drainage system.
A simplified model of the tube Manchette (TAM),
as shown in Figure 4, runs across the diameter of the
cylindrical container. The tube has an internal diameter
of 22 mm, with 4 equally spaced 7 mm holes at the centre. A rubber sleeve covers the holes and 2 rings, one
on either side, prevent the injected grout from flowing
along the tube.
Figure 5. Location of instruments with respect to the injection tube (inj). P are the pore pressure transducers,V measures
the vertical pressure and H the horizontal (Bezuijen et al.
2007).
Changes in pore water pressure during injection

were monitored by four pore pressure transducers,
distributed around the injection point as shown in Figure 5. The readings from these transducers were not
used in this paper. Two total stress cells were used to
record the change in horizontal and vertical pressures.
Two types of sand were used: Baskarp sand
(d50 = 130 m) and Leighton Buzzard type D sand
(d50 = 234 m). In both cases, sand was wet-pluviated
into water in the model container. Loose sand was then
densified to required relative density (70%) by dropping the whole container over 25 mm as many times
as required.
In order to raise K0 to a closer value to 1, sand
was pre-stressed by applying a confining pressure
of 300 kPa at the beginning. Confinement was reduced
to 100 kPa prior to grout injection. This pre-stressing
283
Table 1.
Summary of conducted experiments.
No
W/C
ratio
Rate
(l/m)
Bentonite
(%)
Materials/
sand
5.0
10.0
7.0
GD/Ba
5.0
2.0
7.0
GD/Ba
5.0
10.0
7.0
GD/Ba
5.0
10.0
7.0
Ca/Ba
1.0
2.0
4.0
Ca/LB
1.0
10.0
4.0
Ca/LB
Remarks
Repeatability
test
Slower inj.
rate
Wall friction
effect
Effect of
materials
Lower w/c
ratio
Faster inj. rate
Figure 6. Change of injection pressure with injected volume

for different tests.
Notes: Rate = injection rate, GD = GeoDelft cement

and bentonite, Ca = Cambridge cement and bentonite,
Ba = Baskarp sand, LB = Leighton Buzzard sand. Bentonite
percentage is by weight of mixing water.
was only partially successful and values of starting K0

were still less than, but close to, 1.
Grout injection was conducted using a plunger
pump. A bladder (not shown in Fig. 3) was used as
an interface between pumped water and injected grout
in order to avoid damaging the pump by the grout. The
injection pump was capable of reaching a maximum
pressure of 4 MPa. Injected grout was allowed to set
for 24 hours before the sand was dug out and the shape
of hardened grout was photographed.
4
RESULTS & DISCUSSION
Using the same injection setup, Sanders (2007) and

Bezuijen & van Tol (2007) reported that the best
grouting efficiency was attained by using a cementbentonite grout of a w/c ratio of 5.0 to fracture Baskarp
sand. Ordinary Portland cement and sodium-activated
bentonite (7% by weight of mixing water) were utilized. Injection was made under an injection rate of 10
liters per minute (l/m).
The current series of tests adopted the above mentioned test as a reference and Table 1 summarizes
some of the experiments carried out. Two sets of
grout materials were used: (a) rapid hardening, ordinary Portland cement and sodium-activated bentonite
(GeoDelft, the Netherlands), and (b) normal hardening ordinary Portland cement and sodium bentonite
(University of Cambridge, UK). Injection pressures
resulting from all the tests are shown in Figure 6.
4.1
Repeatability check
Due to a problem with sample preparation that yielded

a sample which was not perfectly homogeneous, the
repeatability test (Test 1, using grout with w/c ratio of
5.0) resulted in a single fracture which propagated to
Figure 7. Fractures from (a) reference test (left) and (b) Test
2 with slower injection rate (right).
near the container wall. Nevertheless, the initial pressure was similar in magnitude to the starting pressure
of the reference test (2.7 and 2.8 MPa respectively).
The pressure during injection was about 40% less than
the injection pressure during the reference test, which
highlights the effect of soil inhomogeneity.
4.2 Effect of injection rate
Reducing the injection rate in Test 2 by a factor of 5 still
yielded fracturing of the sand model. The grout mix
contained enough water and fine particles. Recorded
injection pressure was about 10% lower than the value
for faster injection rate. Sectioning of hardened grout
revealed that thicker fractures were formed, with narrower leak-off zone and thicker filter cake, as shown
in Figure 7b.
Under the slower injection rate, there is more time
for the filter cake to develop. According to Bezuijen
et al. (2007), the thickness of filter cake increases with
the square root of the time that the grout is pressurized.
For a given injection pressure and injected volume,
reducing the injection rate by a factor of 5 will lead to
approximately 2 times thicker filter cake. Formation
of a thicker filter cake will hamper further leak-off, as
the finer particles are blocked by the filter cake that is
already formed.
284
Figure 9. Dehydrated layer around the boundary of hardened grout for (a) Test 6 (left) and (b) Test 7 (right).
Figure 8. Comparison of the change in vertical pressure
with injected volume between the reference test and the tests
with reduced sample height.
4.3
Side wall friction
The sample height for the rest of the tests was 600 mm.
Test 4 was a repetition of the reference test under the
new sample height. Results showed that the injection
pressure under the new testing condition was about
1 MPa lower, even though the injected grout managed
to fracture the sand in a similar way. The change in
overburden pressure that corresponds to changing the
height of sand above the injection point is only a few
kilo Pascals. Therefore, it could not have been the reason for this reduction in injection pressure. However,
measurements of total vertical stress showed that the
friction between the sand and the walls of the container
have reduced the vertical stress at the injection level
from that applied at the top of the sample (Fig. 8). This
explains the reduction in recorded injection pressures
for all the tests conducted under the new sample height.
4.4
Grout materials
The ordinary Portland cement used in the experiments

was CEM I cement and either sodium bentonite or
sodium-activated bentonite was mixed. Bruce et al.
(1997) reported that sodium bentonite is the best type
of bentonites to be added to cement grouts. This is
mainly attributed to its swelling potential, as it swells
up to 18 times its original volume. Sodium-activated
bentonite, on the other hand, could swell up to 10 to
15 times.
Results showed that there was no significant difference resulting from using sodium bentonite instead
of sodium-activated bentonite in terms of fracturing
sand. Comparing the injection pressures of tests 4
and 5 (Fig. 6), the only difference is the slightly slower
build-up of pressure in case of sodium bentonite. This
should have resulted from a problem with the injection
system at the beginning of injection, as almost no heave
or drainage was recorded over the delay period. The
hardening speed of the used Portland cement (rapid
for GeoDelft and normal for Cambridge) did not affect

the results, as most of the processes that influence sand
fracturing happen well before hardening. In terms of
grouting efficiency (defined as heaved volume divided
by injected volume), both types of bentonite gave more
or the less the same efficiency.
4.5 No-fracture tests
Two injection tests were conducted using a low w/c
ratio grout (w/c ratio of 1.0, 4% sodium bentonite
added; see Tests 6 and 7 in Table 1). The tests were
carried out in sand models of type D Leighton Buzzard
sand (d50 = 234 m).
Injection under both slow (Test 6) and fast injection rates (Test 7) yielded no fracturing. The injection
pressures were very close to each other, but slightly
higher than the values for fracturing experiments. The
slower injection rate gave more uniform shape of hardened grout (Fig. 9a), whereas some fingering was
observed for the faster injection rate (Fig. 9b). In both
cases, sectioning the hardened grout revealed a layer of
dehydrated material around the grout-soil boundary.
With higher cement content and less water in the
grout mix, no leak-off occurred and there was limited amount of free bentonite to develop a filter cake.
Bleeding did occur and this in turn reduced the grout
mobility. It is possible that the calcium in the cement
changes the coagulation structure of the bentonite by
cation exchange, increasing the permeability of the
consolidated grout and accelerating bleeding (Sanders
2007). Most of the grout stayed around the injection
point. The existence of a dehydrated layer at the groutsoil boundaries suggests that more bleeding happened
around the boundaries, which is in agreement with the
theory suggested by McKinley & Bolton (1999). The
faster injection rate allowed less time for bleeding and
hence, the grout managed to create some fingering
before it became too viscous to flow.
Consolidation of the grout leads to the possible local
irregularities at the boundaries of an injection hole to
be filled up, or plastered, in the way shown in Figure 10.
285
Figure 10. Influence of plastering on fracture initiation

(Bezuijen & van Tol 2007).
Such plastering will hamper fracture initiation, as it

prevents the fluid pressure from penetrating into the
space between sand particles. As the created plaster
will have a certain strength, part of the injection pressure will be acting on the plaster rather than being
used to push particles away from each other to initiate
a fracture.
5
CONCLUSIONS
The experimental work conducted confirmed that

fracture initiation in sand requires some local inhomogeneity around the injection point, rapid development
of a filter cake with a limited thickness and a grout
with low viscosity and a limited yield stress. Whether
the initiated fractures will propagate or not depends
mainly on the grout mixture. Water-cement (w/c) ratio
and fine particles content play a major role in fracturing of sand. Grouts with high w/c ratios and enough
fines will exhibit a leak-off of fine particles, accompanied with the formation of a filter cake, which results
in fracturing. For grouts with low w/c ratios and large
grout permeability, bleeding is the dominant process,
leading to non-fracturing of sand.
For a suitable grout mixture, faster injection rates
will result in thinner fractures, whereas slower rates
give thicker fractures with less leak-off and thicker
filter cake. If the w/c ratio is too low, no fractures will
be formed, regardless of the injection rate.
The longer fractures experienced under relatively
low injection pressures in the field are mainly due to
the natural inhomogeneity of sand layers. Injection in
almost perfectly homogeneous sand in the laboratory
gives shorter fractures and requires higher injection
pressures.
REFERENCES
Andersen, K.H., Rawlings, C.G., Lunne, T.A. & Trond, H.
1994. Estimation of hydraulic fracture pressure in clay.
Journal of Canadian Geotechnical 31: 817828.
Bezuijen, A., Pruiksma J.P. & Pater C.J.de 2002. Maximum

pressures in tunnelling limited by hydraulic fractures.
Proc. International Tunnelling Association Conference,
Amsterdam.
Bezuijen, A., Sanders, M.P.M., Hamer, D. & Tol, A.F.van
2007. Laboratory tests on compensation grouting, the
influence of grout bleeding. Proc. World Tunnel Congress,
Prague.
Bezuijen, A. & Tol, A.F.van 2007. Compensation grouting in
sand, fractures and compaction. Proc. XIV European Conference on Soil Mechanics & Geotechnical Engineering,
Madrid.
Bezuijen, A. & Tol, A.F.van 2008. Mechanisms that determine between fracture grouting and compaction grouting
in sand. Proc. 6th International Symposium on Geotechnical. Aspects of Underground Construction in Soft Ground,
Shanghai.
Bohloli B. & Pater C.J.de 2006, Experimental study on
hydraulic fracturing of soft rocks: Influence of fluid rheology and confining stress. Journal of Petroleum Science
and Engineering 53 (12): 112.
Bruce, D.A., Littlejohn, G.S. & Naudts, A.M.C. 1997. Materials for ground treatment- A practitioners guide. Proc.
Grouting: compaction, remediation & testing, Logan.
Chang, H. 2004. Hydraulic fractures in particulate materials.
PhD. thesis. Georgia Institute of Technology.
Chin, C.Y. & Bolton, M.D. 1999. Factors influencing
hydrofracture in clay. Proc. 13th ASCE Engineering
Mechanics Conference, Baltimore.
Gafar, K. & Soga, K. 2006. Fundamental investigation of
soil-grout interaction in sandy soils. Report. University
of Cambridge.
Jaworski, G.W., Seed, H.B & Duncan, J.M. 1981. Laboratory
study of hydraulic fracturing. Journal of the Geotechnical
Engineering Division 107(6): 713732.
Khodaverdian, M.M. & McElfresh, P.M. 2000. Hydraulic
fracture stimulation in poorly consolidated sand: mechanism and consequences. SPE Annual Technical Conference, Dallas.
Kleinlugtenbelt, R. 2005. Compensation grouting, laboratory
tests in sand. MSc thesis. Delft University of Technology.
McKinley J.D. and Bolton M.D. 1999. A geotechnical
description of fresh cement grout Filtration and consolidation behaviour. Magazine of Concrete Research 51(5):
295307.
Mori, A. & Tamura, M. 1987. Hydro-fracturing pressure
of cohesive soil. Journal of the soil and foundations,
Japanese Society of Soil Mechanics & Foundation Engineering 27(1): 1422.
Sanders, M.P.M 2007. Hydraulic fracture grouting, laboratory tests in sand. MSc thesis. Delft University of
Technology.
Soga, K., Gafar, K.O., Ng, M.Y.A. & Au, S.K.A. 2006. Macro
and micro behaviour of soil fracturing. Proc. International Symposium on Geomechanics and Geotechnics of
Particulate Media, Yamaguchi.
Soga, K., Ng, M.Y.A. & Gafar, K. 2005. Soil fracturing in grouting. Proc. 11th International Conference of
the International Association of Computer Methods and
Advances in Geomechanics, Tornio.
Thallak, S. 1991. Numerical simulation of hydraulic fracturing in granular media. Phd thesis. University of Waterloo.
286
Historical cases and use of horizontal jet grouting solutions with 360
distribution and frontal septum to consolidate very weak and saturated soils
G. Guatteri
Novatecna Consolidaes e Construes S.A So Paulo, Brazil Terrajato Lisbon, Portugal
A. Koshima, R. Lopes & A. Ravaglia

Novatecna Consolidaes e Construes S.A So Paulo, Brazil
M.R. Pieroni
Novatecna Consolidaes e Construes S.A So Paulo, Brazil Terrajato Lisbon, Portugal
ABSTRACT: Horizontal jet grouting technology has proved to be quite a versatile tool for dealing with the
particular geological conditions encountered when excavating tunnels in soil mass, especially in the presence of
water flow and high hydraulic gradient. We describe the state-of-the-art application known as 360 distribution,
in which horizontal jet grouting columns are executed around the excavated section, including the invert, and at
the far end of the conical treatment, to create a watertight chamber. This results . in a heading that is constantly
protected by pre-consolidated soil and minimizes the effects of excavation on nearby structures. We also describe
a special tool (called the preventer) designed to control drilling and injection fluid outflow, which proved to be
absolutely essential to successful outcomes. We also present the experience gained by the authors in a number
of projects in Brazil and Venezuela and the results of a full scale test, executed for the first time in Europe.
INTRODUCTION
Horizontal jet grouting has proved to be an efficient

and versatile technical solution and is increasingly
being used for tunnelling in difficult soils, usually with
high granulometry, presence of intense water flow and
high hydraulic gradient.
Further developments have facilitated more nonTBM excavations in locations with low overburden
and superstructure presence, such as urban areas
where stable and safe conditions during excavation are
required at all times, with minimum impact on nearby
structures.
One type of application that is not yet in common
use is the one in which horizontal jet grouting columns
are executed all around the tunnel cavity in what is
called a 360 distribution (roof, sidewalls and invert)
and in which, depending on the circumstances, consolidation is extended to form a horizontal soil-cement
frontal wall (septum) at the far end of the conical
treatment.
Therefore, for the heading of the tunnel, a sequence
of watertight chambers are created so that the excavation activity is always protected by pre-consolidated
soil.
We provide a brief description of this innovative

technical solution, its main geometrical characteristics
and use conditions.
From the executive point of view, we describe the
use of a special tool named the Preventer, which was
developed to control the outflow of drilling and injection fluids, and is absolutely essential in order to obtain
good performance.
Finally practical cases, executed by the authors,
are described, along with the results of the first full
scale test of a 360 jet grouting treatment executed in
Europe.
2
360 DESIGN
In recent years, horizontal jet grouting technology has

frequently been used for tunnels where safety was a
priority and more traditional techniques were not feasible or might not guarantee good performance. As
readers will know, this applies mainly to consolidating soil mass with poor geo-mechanical characteristics
which do not allow excavation in safe conditions,
mainly in sandy, silty or clayly soils or a combination
of these, with or without presence of water.
287
Figure 3. Example of 360 distribution for half-section and

bench heading.
Figure 1. Typical distribution for roof and lateral protection
and forepoled nucleus.
Figure 4. 3D Analysis of a 360 distribution.

Figure 2. Typical distribution of a full face 360 distribution.
This technology is normally associated with the

NATM . In a typical configuration, columns are executed side by side to form a pre-consolidated arch
which follows the external profile of the excavated
section.
Depending on the actual soil conditions, the consolidation effect may be improved by jet grouted columns
executed inside the section to reinforce the tunnels
nucleus, or by means of horizontal drains drilled at
the bottom or around the section. In more severe situations, the jet grouted columns may be reinforced by
means of steel pipes or fiberglass, and drains may be
equipped with vacuum systems. Fig. 1 shows a typical
traditional distribution.
The encouraging results obtained from the horizontal jet grouting technique and the growing needs for
executing tunnels in increasingly difficult conditions,
at shallower depths, generally in alluvial or even residual deposits but with high granulometry, intense water
flow, hydraulic gradient and the presence of important
superstructures, as in many urban projects, prompted
the development of the 360 distribution.
The main characteristics of the 360 solution, compared to the traditional, are: a) soil consolidation, previously limited to roof and sidewalls, is now extended
to the invert, and b) the execution of a jet grouted plug
at the bottom of the conical treatment.
Thus a sealed chamber can be created, and in terms
of stability, soil-mass loads around the cavity are redistributed to prevent possible inflows of material from
the excavation face and the invert, which are quite

common in sandy soils.
From the hydraulic point of view, the drastic reduction of water inflow during the excavation phase
reduces soil-mass dewatering and therefore keeps settlement under control and minimizes effects on nearby
structures while the heading is executed. Fig. 2 shows
a typical example of a 360 degrees distribution with
full-face excavation.
Depending on the tunnel dimensions, the solution
may also be adjusted for use with a half-section heading. In this case, the bench or final invert may be
executed, if required by the geological conditions,
with the help of vertical or inclined jet grouting to
ensure safer conditions and optimize working schedule
(fig.3).
As far as the operational aspect is concerned, the
360 solution became feasible due to advances in
drilling equipment, involving redesign and improvement of its maneuverability, set up and mast alignment.
Special homothetic templates with the projection of
the consolidation elements for the perfect alignment
of the drilling string had to be developed and implemented.
Also, 3D graphic tools were introduced in order
to verify the integrity and continuity of the treatment
designed (fig.4).
Special consideration must be given to the device
used to control the outflow of drilling and injection
fluids , which evolved from a similar one used in the oil
industry, hence the name preventer. Proper use of this
device proved to be essential to achieve appropriate
288
Figure 6. Tunnel face with starting drilling points.

Figure 5. Preventer typical configuration.
monitoring and control of movement of excavated soil

mass.
3 THE PREVENTER CONTROL VALVE

The preventer valve (or blow-up preventer) is used
to avoid gas leaks and explosions when drilling oil
wells.
Preventer valves and retainers at the mouth of the
hole control outflows of solid and liquid spoils from
drilling when ground is sandy and susceptible to piping. They also facilitate control of slurry (excess
soil-cement mixture) during jet grouting jobs, thus
avoiding piping along the hole or the column.
Note that a drilling operation in non-cohesive soils
below the water table produces withdrawal of material
due to the intense flow established by the water table
hydraulic gradient and water needed for the drilling
operation itself.
In such case, it quite common to lose control of the
volume of material flowing out of the hole if there is not
proper perception of the phenomenon, thus triggering
a piping effect.
This piping occurs when pressure inside the hole
during the drilling operation is less than the effective
hydrostatic pressure acting into the ground. And this
can occur also during the jet grouting phase.
In this situation, with the preventer installed at
the hole mouth, we may control pressure inside the
hole, thus avoiding decompression of the surrounding
material leading to an increase of permeability.
This device (fig.5) has a system of valves and seals
to maintain strict control of volumes of materials withdrawn during drilling or injection. At the end of the
jetting operation, it can also be sealed for the time
needed for the soil-cement mixture to set.
Figure 7. Septum and installation of yielding arch.
HISTORICAL CASES
4.1 Copacabana subway tunnel (Rio de Janeiro

Brazil)
Rio de Janeiros subway company planned a 750 m
extension to line 1, from Cardinal Arcoverde Station to
the center of Copacabana, also building Siqueira Campos Station and turn-offs or maneuvering areas. The
underground structure comprised two independent but
juxtaposed eyeglass type tunnels crossing gneissic
rock, saprolite and micacious residual silty-sandy soil
and sandy clayish soil of marine origin, of high hydrogeological complexity; the route also crossed a densely
populated built-up area.
One of the greatest challenges was the NATM section of the tunnels under the direct foundations of
a 50-year-old 7-storey building, standing on sandy
sediment (beach sand), with around 6m overburden.
These tunnels were previously treated with horizontal jet grouting forming a closed 360 conical chamber
with treatment throughout the excavation cross section and front septum protecting each advance module.
(figs. 6, 7).
289
Figure 10. 360 horizontal jet grouting scheme.
Figure 8. Plaza Italia/Capuchinos Twin Tunnels.
Figure 11. Geological profile and tunnel path details.
Figure 9. Section details.
The work was successfully concluded on time and

the preventer valve was a crucial factor in ground
treatment. The maximum settlement recorded was
30 mm.
In order to cope with these difficulties, various

preventive measures were taken in the design and execution phases. First of all, the design was adjusted to
the in situ conditions, extending the vault treatment
to the invert or locally, whenever it was necessary,
so as to pre-consolidate and reduce water flow (360
Horizontal Jet Grouting, fig. 10).
Due to the presence of water under pressure, the
drilling and jet grouting operations at the invert had to
be executed using the preventer valve.
Since the central portion of the section proved to
be in clay material, there was no need to execute a
jet-grouted frontal septum.
However, specific measures to control pressure and
water flow using deep horizontal drains, equipped with
check valves when necessary, had to be taken.
Plaza Italia Tunnel Line 4 subway (Caracas

Venezuela)
4.3 Tamanduatei river service tunnel (So Paulo

Brazil)
Part of the new Line 4, the North and South Tunnels Initial Sections, heading from Plaza Itlia toward
Capuchinos Station, involves the expansion of an
existing station where two tunnels of approximately
6.5 m diameter were to be excavated on each side of
an existing tunnel, which had to be maintained operational (fig. 8 and 9). In this case the existing structure,
acted as a barrier to the ground water flow, raising its
level and pressure.
On starting the excavation, instead of the residual
soil predicted by the original geological investigation,
tunnelers found highly permeable alluvial sandy soil
with gravel at the bottom of the excavation section,
topped by a clayly layer of residual origin (micaceous
schist), and a sedimentary fine sandy layer, at the
upper part of the tunnel. The overburden was around
10 meters.
To our knowledge, this is the first case worldwide in

which the 360 distribution has been adopted as a pre
consolidation shape in a NATM tunnel.
In this case the challenge was the tunnel running at
about 10 m under street level, but deepening to 25 m
when crossing the river (Fig. 11) due to the project
requirement for a minimum of 5 m under the bottom of
the deepest diaphragm-walls containing the canalized
river; this structure would pose a serious risk of becoming a preferential water-path communicating with the
river bed.
The geological profiles (Fig. 11) showed that the
tunnel is embedded in an alluvial mass deposited
during the more recent Tertiary period. The matrix
comprises coarse, medium and fine light-gray sands,
with N(SPT) 10 to 40 and is very pervious at the
tunnels levels.
4.2
290
Figure 13. Full Scale Test General lay-out.
Figure 12. Treatment scheme with full section septum.
Inside this matrix, two highly plastic silty clay layers, 2 to 5 m thick, are intercalated but only randomly
intersecting the tunnel excavation.
Water table observations showed a quick response
during rainy periods. The same effect was produced
whenever the level of the river rose due to rainfall
anywhere along its upper course.
Before the excavation, as additional precaution,
watertightness was checked by installing three horizontal drains inside the chamber so as to verify the
presence of water under pressure and/or continuous
flow.
4.4
Full scale test for a railroad tunnel (Barcelona

Spain)
The Madrid-Zaragoza-Barcelona-French Frontier high

speed rail links section connecting to Sants Station in
Barcelona, required a tunnel with a 120 m2 cross section located underneath the path of the existing ground
level railway.
The geological investigation detected the presence
of alluvial sandy material with high water table and
high permeability, which was potentially dangerous
for the tunnels stability and the existing nearby structures, without special soil consolidation treatment.
In order to verify the efficiency of the horizontal
or inclined jet grouting consolidation techniques in
submerged conditions, it was decided to carry out a full
scale test to confirm the feasibility of some alternative
technical solutions proposed, verify the consolidation
design parameters and check the effect of the activity
at ground level, recording settlements and water table
variations.
Figure 14. Geological Profile related to the chamber position.
On a slightly smaller scale (80% of the original dimensions), the test reproduced the executive
conditions of a consolidated horizontal half-section
heading with all the structural and waterproof elements executed in accordance with the procedures and
specifications stipulated for the actual structure.
Fig. 13 shows the general lay-out and geometry
of the proposed treatment. The heading has been
dimensioned to simulate the excavation of a stretch
of tunnel 5 meters long between two jet grouted septums. The soil cover above the tunnels crown is around
6 meters. Also the logistic and operational conditions
facing the excavation crew during the actual work were
accurately reproduced.
The geological profile at the test area is shown in
fig.14. Basically it consists of a sequence of sandy to
silty soils with SPT varying from 5 to 15, intercalated
with very thin layers of silty clay. The presence of stiff
gray clay was detected only well below the invert.
Between the tunnels crown and invert levels, the
soil is a very fine silt that is under hydraulic gradient and flowing out of the excavation almost without
control. Since above the crown level there is no more
291
Figure 15. Instrumentation lay-out.
Figure 17. Piezometers Readings.
Figure 16. Bench Mark HN 40 Readings.
consistent strata, even a small outflow is immediately

reflected at the surface.
To verify the effects of the jet grouting injection onto
nearby structures and the efficiency of the chambers
watertightness, piezometers and bench marks were
installed, properly distributed above and around the
test area (see fig. 15).
A number of horizontal drains with vacuum system
were installed inside the excavation section in order
to facilitate the chamber dewatering. A connection
between the chamber and the soil mass outside would
be detected by the external piezometers. The readings
of settlements and water table levels were taken daily.
Fig. 16 and 17 show a summary of the data recorded
throughout the test period. Despite very poor soil characteristics, it was possible to keep soil movements
around the chamber and at ground level under control,
with an average variation of about +/20 mm.
The preventer (control valve system) was permanently installed except in a particular situation when,
due to a technical problem with this device, a massive
loss of material was experienced. Once the preventer was repaired, the original ground conditions were
reinstated by injecting and recompressing the ground
area affected. A pressure gauge was installed at the
mouth of the device to verify the build-up of pressure
inside the soil mass. A maximum value of 0.4 Mpa was
recorded.
Piezometers readings were stable during the jet
grouting consolidation phase. During the chamber
excavation, one piezometer showed an anomaly indicating a possible gap in the consolidated area.
The problem was solved by reinforcing the consolidation with a small number of additional columns on
one side of the external alignment limited to the area

affected by the phenomenon. The chamber excavation
was successfully concluded confirming the stability
and watertightness of the proposed treatment.
The test results were fully adopted in the design
and subsequent execution of the main tunnel under
the existing railway.
CONCLUSIONS
The 360 consolidation treatment is the state-of-theart in terms of pre-consolidation technique associated
with NATM mainly in alluvial or residual formations,
with high granulometry and relevant presence of water
pressure, and in general in all types of soil were jet
grouting technique can form the column element.
To ensure the level of efficiency required, the
instruments to pursue treatments imperviousness and
geometry are the preventer valve that is essential in
order to avoid uncontrolled loss of material due to piping together with the special homothetic templates for
perfect alignment of the drilling string.
The 360 consolidation scheme proved to be
extremely flexible, as it can be adjusted to actual in situ
geological conditions, and when executed in the appropriate manner and with proper equipment, achieves
extremely good results in terms of ground movements,
with a recorded average of 20 to 30 mm thus allowing
a proper control of side effects on existing structures.
So far this technique has be implemented in
urban tunnels with a minimum overburden of around
6 meters. The encouraging experience gained since
1998 when was executed for the first time, is at present
leading the authors to study its application in even
shallower tunnels with a soil cover of 3 to 4 meters.
292
REFERENCES
Guatteri, G. Mosiici, P. Doro Altan, V. Koshima, A. 1998.
Brazilian Experience in Jet Grouting Treatments in Difficult Tunnels. World Tunnel Congress 98, ITA CBT,
Tunnels and Metropolises, So Paulo.
Guatteri, G. Koshima, A. Lopes, J. Doro Altan, V. 2000.
360 Jet Grouted Conical Chamber Allow Safe Tunneling
Under River Within a Highly Pervious Environment. Geoenge 2000 An International Conference on Geotechnical
& Geological Engineering; Melbourne, Australia.
Guatteri, G. Koshima, A. Lopes, J. Pieroni, M. 2004. Resume

of Brazilian Experience about Use of Jet Grouting in Tunnels and Underground Excavations. 1 Brazilian Congress
of Tunnels So Paulo.
Guatteri, G. Koshima, A. Ravaglia, A. Duarte, M. 2007.
Jet Grouting Solutions for the Caracas Subway Line 4.
Challenges in Urban Projects XIII Panamerican Conference on Soil Mechanics and Geotechnical Engineering
Venezuela
Szio, L. 2002. Jet Setting in Rio. Tunnels & Tunneling
International, I-1,15.
293
The effects of sample dimension and gradation on shear strength parameters

of conditioned soils in EPBM
M. Hajialilue-Bonab, M. Ahmadi-adli, H. Sabetamal & H. Katebi
University of Tabriz, Tabriz, Iran
ABSTRACT: Mechanical properties of conditioned soils in EPBM tunneling consist of lots of unknowns.
In this research, the tests has been arranged to fulfill of four goals. Firstly the effects of conditioning on the
shear strength variation have been investigated. Secondly an investigation on effects of conditioning on shear
strength parameters (C,) has been performed. In third step the results exerted from two previous stages have
been compared for two shearing apparatuses of conventional and large shear boxes. The last goal of the research
is exploration of effects of changes in conditioning parameters on shear strength. It is found that the C& for
tested soils obtained from large shear box are usually greater than the results of the same soil in conventional
shear test. This result is less significant for conditioned soil with compared to unconditioned soil and it is found
to be a function of injected foam content.
INTRODUCTION
The last parameter is of great importance in successful soil conditioning.
Determining the shear strength parameters of the conditioned soils usually lead to the use of conventional
shear box tests because of no availability to large shear
boxes. In order to use this apparatus, soil gradation
must be modified based on the codes which demands
to elimination of great mass of coarse particles. This
elimination affects the shear strength parameters especially in conditioned soil. In this research the effects of
sample dimension and soil gradation on properties of
foam conditioned soils have been investigated. A modification on soil gradation has been done in samples
of large and conventional shear boxes. The variables in
this research are normal stress and the injected foam
quantity in to each of soils.
2
2.1
MATERIALS AND PROCEDURES IN SOIL

CONDITIONING
Conditioning parameters
Concentration (Cf )
This parameter is the content of foaming agent in
water unit weight.
Foam Expansion Ratio (FER)
This is the volumetric expansion of a unit volume
of foaming solution.
Foam Injection Ratio (FIR)
It consists of the ratio of injected foam volume to
volume of the conditioned soil.
2.2 Laboratory foam generation system

To produce a conditioner (foam) with specific and
controlled parameters, a foam generating system in
laboratory scale was designed and constructed. This
apparatus has the ability of controlling foam properties
along the production and can produce uniform foam.
The schematic plan of this system which manufactured in soil mechanic laboratory of Tabriz University
is shown in figure 1.
Foam generation is performed by mixing process
of air and liquid (consist of foam agent) under pressure. In the first step the reservoir must be filled by
foam Solution. Foam generation is performed by mixing process of air and liquid (consist of foam agent)
under pressure. In the first step the reservoir must
be filled by foam solution. A regulator controls the
air pressure supplied to the reservoir. Liquid valve is
opened and flow control valve is used to adjust the
flow through the liquid flow meter to the foam generator unit inlet. The pressure in the air flow line is
controlled with a control valve of liquid flow meter
and measured by a pressure gauge. Air pipe valve
is opened to allow the air to flow through the control valve and the air flow meter to the generator unit
inlet. The pressurized air and foam solution then flow
through the generator unit to produce the foam. The
design of the foam generator allows the liquid and
295
1.2
shear/normal stress
a
1
0.8
0.6
0.4
Unconditioned
0.2
Conditioned
0
0
10
Horizontal disp(mm)
1
b
Vertical disp(mm)
0.8
Figure 1. Schematic of laboratory foam generation system.

100
0.4
0.2
Unconditioned
Conditioned
0
-0.2
90
-0.4
80
-0.6
0
70
10
Horizontal disp(mm)
60
%
0.6
50
pA
40
30
A'
20
C'
Figure 3. (a) Variation in ratios of shear to normal stress

and (b) vertical displacement of sample cap versus horizontal
displacement in Conventional shear boxes on soil A under
normal stress 37.52 kPa.
10
0
0.01
0.1
10
100
mm
4.1 Conditioning effects on shear strength

variation trends
Figure 2. Particle distribution of soils used in this research.
air flow rates and pressures to be adjusted and monitored separately to control the properties of the foam
produced.
RESULTS AND DISCUSSIONS
SHEAR TESTS PATTERNS
In order to perform large shear box (300 300

150 mm) tests and conventional shear box (60 60
20 mm) tests the soil excavated from Tabriz metro line
has been used. This soil labeled pA. Considering the
maximum dimension of the soil particles with respect
to the shear box dimension (large, conventional), soil
gradation has been modified to A, A (by elimination
of the oversized particles) and C (same as A with
30% more in fine content). Particle size distribution for
these soils has been illustrated in figure 2. Direct shear
tests on soils A, A and C were performed using conventional and large shear boxes. The normal stresses
on samples also were 37.52, 64.75, 119.29 and 228.28
(kPa).
The behavior of soil A in conditioned and unconditioned state under different normal stress (37.52, 64.75
and 119.29 kPa) in direct shear tests have been shown
in Figure 3 to 5. The variation of shear/normal stress
ratio versus horizontal displacement has been investigated. The variation of vertical displacement with
respect to horizontal displacement during shear test
has been also considered. Figure 3 shows the results
for soil Ain both conditioned and unconditioned state.
A reduction of almost 0.4 in maximum values of stress
ratios can be observed. The curvatures indicate some
peak point in lower stress ratios in unconditioned soil
test. This peak is going to disappear by increasing
the stress level. But in conditioned samples there is
no peak point in any stress level. It can be observed
that increasing in stress level can convert the dilative
behavior to contractive behavior. However the conditioned samples give a completely contractive behavior
regardless of initial specific volume.
4.2 Conditioning Effects on shear strength
parameters
Table 1 shows calculated C& for performed tests. The
friction angle for peak and residual states are given.
296
1.2
1
shear/normal stress
shear/normal stress
0.8
0.6
0.4
Unconditioned
Conditioned
0.2
0
0
10
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
Unconditioned
Conditioned
Horizontal disp(mm)
0.8
b
0.7
vertical disp(mm)
0.6
0.4
0.2
Unconditioned
Conditioned
10
Horizontal disp(mm)
0.8
vertical disp(mm)
-0.2
0.6
0.5
0.4
0.3
0.2
Unconditioned
0.1
Conditioned
-0.4
0
10
10
Horizontal disp(mm)
Horizontal disp(mm)

The residual friction angle represents the state of the

soils in excavation process.
Comparison of corresponding shear tests performed on both conventional and large shear boxes
indicate that the large boxes give internal friction
angle 6 to 7 degrees greater than conventional shear
tests. However in the tests on conditioned samples this
reduces to 3 to 4. Large apparatuses also result in lower
cohesion in comparison to the standard samples. Also
in conditioned samples this trend is traced but it is less
intensive.
4.3 Effects of variation in conditioning parameters

on shear strength parameters
The needed conditioner for soils used in tests has been
predicted using the Kusakabe 1999 formula.
Where X = the percentage of soil passing 0.074 mm;

Y = the percentage of soil passing 0.25 mm; Z = the
percentage of soil passing 2.0 mm; a = 1.0 for Cu > 15,
a = 1.2 for 15 > Cu > 4, a = 1.6 for 4 > Cu

Using this formula for soil A, the FIR is 50%. Soil A

has been tested with FIR equal 50%, 30% and 70%
and results are shown in figure 6. For example in
treatment with FIR = 30% an increasing of 0.11 in
normal/shear stress ratios can be observed. Contrarily,
with FIR = 70% no remarkable effect was observed.
It is clear that increasing FIR do not change the normal/shear stress ratios while a little reduction on FIR
has considerable effect on this ratio.
Figure 7 gives the results for another group of similar tests on soil A with various FIR. If one apply the
soil gradation A to the Kusakabe formula, FIR will
result in 40%. By increasing foam consumption.
In soil samples, a negligible reduction in stress
ratios can be observed, but a little reduction in foam
consumption will result a significant increase in these
ratios. This relation in conventional shear tests is about
0.17. This means that consuming more foam indicates slight effect on shear strength of samples. During
shearing, pore pressure will generate but in granular
soil this will dissipate very quickly. In the presence
of foam the dissipation of pore pressure will postpone
and the shear strength will decrease in foamed granular soil. Comparison of changes in stress ratios in large
and conventional tests shows that conventional direct
shear tests results are more sensitive to the changes in
297
37.52 45.322
64.75 72.888
119.29 125.08
50 37.52 26.095
64.75 44.93
119.29 75.929
0 37.52 42.281
64.75 63.667
119.29 105.95
228.28 186.68
40 37.52 27.076
64.75 43.066
peak
shear/normal stress
kPa
residual
C
kPa
50.38 9.221
48.39
46.36
34.82 4.42
34.76
32.48
48.41 14.62
44.51
41.61
39.28
35.81 7.259
33.63
0.7
0.6
0.5
0.4
0.3
FIR=40%
FIR=20%
FIR=60%
0.2
0.1
0
44.2
10
Horizontal disp (mm)

0.9
31.1
0.8
vertical disp(mm)
Conventional
shear box
0.9
0.8
Normal stress
Fine content
Large shear box
Shear parameters.
Amount of conditioner
Table 1.
37.4
28.4
0.7
0.6
0.5
0.4
0.3
0.2
FIR=40%
FIR=20%
0.1
FIR=60%
shear/normal stress
0
0
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
10

FIR=50%
FIR=30%
FIR=70%
60
50
0.5
1.5
2.5
3.5
40
Horizontal disp (cm)
Figure 6. Variation in ratios of shear to normal stress versus

horizontal displacement in large shear boxes on soil A under
FIR than the results in the same samples in large shear

test. The results for other tests give the same conclusion. This can be observed in Figure 8. This figure give
some more information about soil C which is similar
to soil Awith some more fine content. Its obvious that
the increasing FIR in samples that have excessive fine
causes more significant loss in internal friction angle.
But in samples which have more fines this phenomena firstly occurs steeply but in FIRs higher than
optimum this trend is not observed at all.
It can be concluded that the soil which have more
fine content need less FIR to get a specific reduction
of shear strength.
In order to study the possible effects of variation in
foam expansion ratio (FER), a foam in three FERs of
5, 10 and 22 were prepared for similar shear tests on
30
A
20
A'
10
C'
0
0
10
20
30
40
50
60
70
80
FIR %
Figure 8. Internal friction angle versus variations in FIR.
soil C. Figure 9 shows that a reduction on FER causes

no effect on shear parameters and stress ratios.
The reason which can be stated is that the low FER
indicates very wet foam so that it is unable to reduce the
relative density and consequently increase the specific
volume. On the other hand when the FER increases an
unstable state occurs and consequently no changes in
stress ratios happen. It is resulted that a FER closer to
15 gives a stable state for foam.
Figure 10 gives a view of obtained variations in
stress ratios of conditioned samples during shearing
298
0.9
0.7
a
0.6
0.7
shear/normal stress
shear/normal stress
0.8
0.6
0.5
0.4
0.3
FER=5%
FER=10%
0.2
FER=22%
0.5
0.4
0.3
Cf=3%
Cf=4%
Cf=5%
0.2
0.1
0.1
10
10
0.7
0.7
0.6
Vertical disp(mm)
vertical disp(mm)
0.8
0.8
0.5
0.4
0.3
FER=5%
FER=10%
0.2
FER=22%
0.6
0.5
0.4
0.2
0.1
10b-
Cf=3%
Cf=4%
Cf=5%
0.3
0.1
10
10

displacement in Conventional shear boxes on soil C under
because of changes in Cf. Three tests performed on C

with three different initial Cf of 3%, 4% and 5%. If we
consider the possible changes in stress ratios there is no
significant difference between these samples. But the
volume of samples tends to reduce during the test. The
authors considered this logical as well as declaration
of producers of conditioners.
CONCLUSIONS
1. The soil conditioning cause a decrease of 7 to 11

degrees in internal friction angle of soils. A loss of
49.1 kPa for the cohesion of conditioned soil was
observed.
2. Decreasing FIR from its optimum given by
Kusakabe formula, will greatly affect the trend of
reduction in shear strength. But increasing in that
quantity does not show remarkable effect. FER
lower than 10 and more than 18 causes no conditioning and consequently the changes in strength
are negligible.
3. Large shear test gives greater internal friction angle
about 6 to 7 degrees in unconditioned samples and

displacement in Conventional shear boxes on soil C under
3 to 4 angles in conditioned samples comparing

conventional shear test.
REFERENCES
Ahmadi adli, M, Investigation on mechanical properties
of conditioned soils for EPB mechanized tunneling in
Tabriz, 2006, M.Sc. thesis, Faculty of Civil engineering,
University of Tabriz. Tabriz: Iran.
Defence Standard 4240. 2002. Foam Liquids, Fire Extinguishing (Concentrates, Foam, Fire Extinguishing).
Ministry of Defence. Issue 2. USA.
EFNARC. 2005. Specifications and Guidelines for the Use
of Specialist Products for Mechanized Tunnelling (TBM).
In Soft Ground and Hard Rock. EFNARC, UK.
Merritt, A. 2004. Conditioning of clay soils for tunneling
machine screw conveyors. PH. D. thesis, St. Johns college,
Cambridge university. London: England.
Milligan, G.W.E. 2000. Soil conditioning and lubrication
in tunneling, pipe jacking and micro tunnelling. A
state of art review. http/www.civils.eng.ox.ac.uk/research/
pipejacking.htm
Psomas, S. 2001. Properties of foam/sand Mixtures for tunnelling applications. M.Sc. thesis, St. Hughs college,
Oxford university. London: England.
299
Experimental study on compressibility behavior of foamed sandy soil

M. Hajialilue-Bonab, H. Sabetamal, H. Katebi & M. Ahmadi-adli
University of Tabriz, Tabriz, Iran
ABSTRACT: In order to assess the influence of different foam types on compressibility behavior of conditioned
sand, a set of tests were performed on three gradation of sandy soil. Some index tests were also undertaken for
verifying foam agents characteristics and foam generator quality. Details and discussion about different aspects of
mentioned cases have been presented in this paper. Compressibility tests were performed by a 151 mm diameter
Rowe Cell and foam generation was carried out by foam generator which was constructed by the authors.
INTRODUCTION
Ideal ground conditions for EPB machines consist of

soils with relatively high fines contents such as clayey
silts or silty sands, with a consistency to form a low
permeability, soft plastic paste when excavated. These
properties allow the support pressure to be transferred
uniformly to the tunnel face and controlled flow of
the soil through the machine. The compressibility of
soil has an important function on machine performance. However, natural soils rarely have these ideal
properties, and conditioning of the soil is usually necessary to change its properties to suit the machine.
Soil conditioning for EPB machines involves injecting conditioning agents into the excavated soil. The
objective is to modify the soil properties to form a soft
plastic paste, leading to improvements in the machine
performance and control of the excavation process in
a wide range of soils. The specific treatments required
to effectively condition different types of soil. Many
factors influence the specification and performance of
soil conditioning treatments.
Soil conditioning is clearly performed by means of
some chemical & physical materials such as foam and
polymer which lead to decrease permeability, increase
plasticity, etc. The SCA (Soil Conditioning Agent) is
also responsible for controlling the rheological properties of the extracted soil, and minimizing wear and
abrasion of the cutter head during tunneling. In order
to achieve all of these objectives, the SCA composition must be tailored to the properties of the soil.
Foams, mainly consisting of surfactants, are used in
fine-grained materials to reduce material adhesion to
exposed EPBM surfaces, and to fluidize the muck
(reduce balling and enhance flow ability). Foams are
also used in granular deposits to improve material rheology and reduce soil permeability. While the use of
SCAs in EPBM tunneling is on the rise, the use of

SCAs in general is very much a black box technology. According to Psomas (2001) foam/sand mixtures
exhibit low shear strength and they have very compressible, so in order to assess the influence of foam
and polymer on shear strength and Compressibility
behavior and influence of foam parameter in those,
supplementary tests performed with variety of foam
agents and foam parameters such as FER, FIR and Cf
on three gradation of sandy soil (see Sabetamal 2006).
This paper focused on the results of compressibility
tests which have been undertaken by two types of foam
agents.
2
2.1
EQUIPMENT AND MATERIALS

Foam generator
Lab scale TBM foam generator with full controlling

possibility on various part of system was designed and
constructed. The operation of the laboratory foam generator is described here with reference to the schematic
diagram shown in figure 1. Foam generation is performed by mixing process of air and liquid (consist of
foam agent) under pressure. In the first step the reservoir must be filled by foam solution. Then it flows on
route 2 via air pressure. A regulator (b) controls the air
pressure supplied to the reservoir and the compressed
air is flowed in route (1). The entrance air pressure is
measured by gauge a. Valve (2) is opened and flow
control valve is used to adjust the flow through the liquid flow meter to the foam generator unit inlet. Valve
(1) is opened to allow the air to flow through the control valve and the air flow meter to the generator unit
inlet. The air pressure is measured by a pressure gauge
(e). The pressurized air and foam solution then flow
301
Figure 2. Compressibility tests equipment set up.
Figure 3. Sands gradations.

Table 1.
Figure 1. a: Photo of laboratory foam generator.

b: Schematic of laboratory foam generator.
through the generator unit to produce the foam. The

design of the foam generator allows the liquid and air
flow rates and pressures to be adjusted and monitored
separately to control the properties of the foam.
Soil
GS
emin
emax
D50 mm
A
B
C
2.65
2.65
2.65
0.452
0.660
0.580
0.695
0.940
0.766
1.0
0.25
0.60
pressure are measured using LVDT and pressure

transducer respectively according to figure 2.
3
2.2
Sands properties.
Rowe cell
Compressibility tests were carried out in a standard

151 mm diameter Rowe Cell. Stiff porous discs were
placed on the top and bottom of the sample so that
the boundary conditions are fixed strain rather than
free strain. All tests were undertaken at seven stress
level and each level composed of two stages: undrained
stage (immediate settlement is obtained) and drained
stage (settlement due consolidation is obtained). At
each stress level, the measurement of drained liquid
volumes in second stage leads to calculate the air
and liquid void ratio. Vertical settlement and vertical
STUDIED MATERIALS
3.1 Sand
The gradation curves of different sandy soils which
were used in this research are shown in figure 3. The
soil type A and B are quite uniform. The soil type C is
a mix of soil A and B. The properties of these sandy
soils such as Gs , emin , emax are shown in table 1
3.2 Foam agents
Four different types of foam agent and one type of
polymer were used in foam index tests. Only two of
302
them were used in compressibility tests. Agent type

D is produced in Iran and vastly used in Shiraz metro
project. Other agents (named A, B, C) and polymers
(named P) are produced by Degussa. The foam generator was able to operate with all of them that produce
acceptable quality micro-foam.
FOAM INDEX TESTS
The foam properties depend on its different compounds like air, water, surfactant and sometimes
polymer. The parameters which characterize foam are:
Surfactant Dosage = Cf [%]

Polymer Dosage = CP [%]
Air Ratio (Foam Expansion Ratio) = FER
Foam Injection Ratio = FIR
25% or 50% drainage time according to Minis try
of defense Standard 4240
The amount of air introduced to the soil can be

changed with the air ratio FER which characterizes the
ratio between air and liquid volume. The foam injection ratio FIR indicates the volume of foam used per
1 m3 excavated soil. In order to verifying the properties
of produced foam and to investigate the capability of
foam generator to produce qualified foam with different agents, 50% drainage time were tested for produced
foams.
According to the index tests, it is found that to produce of foam with equal FER, the amount of air and
liquid flows must be adapted with each foam agent
type. The amount of FER with respect to Cf for foam
D at the same condition is shown in figure 4a. It
can be observed that, with increasing of Cf, FER will
increase.
The stability times for different foam are shown in
figure 4b.The stability time varies from 6 to 15 minutes
for different foams. By adding 2.5% polymer to foam
agent B the stability time will be doubled. This can
be observed for all foam agents. This is an important
result as the stability time is very important for some
kind of excavations.
Figure 4. a: Effect of Cf on FER for foam D b: Comparing

of foam stability.
Figure 5. Drained and undrained settlement at end stage

(@450 kPa).
COMPRESSIBILITY TESTS
The compressibility tests were normally performed at

four stress level in which the stress be doubled with
respect to previous level. Because of sample sensitivity
to stress level, the tests were carried out at seven stress
levels. For this reason the incremental ratio decreased
from 2 to 1.414 from tertiary step (i.e. doubling in
two load steps). During each load stage the drainage
valve was closed at the beginning, in order to measure an undrained compression. Such a compression
would be negligible for a saturated specimen, but for
the sand/foam mixtures there is a significant compression of the air bubbles in the foam. Then the drainage
valve was opened and fluid allowed to drain from sample. A typical plot of displacement against time which
shows both the undrained and drained phases is given
in figure 5.
Figures 6, 7 illustrate the variation of void ratio
against vertical stress in semi logarithmic scale for
303
Figure 6. Void ratio against load logarithm for sand B.
Figure 7. Void ratio against load logarithm for sand C.
sands B, C. Only the last points for each load increment are plotted, i.e. the void ratio achieved at the end
of the drained stage. The soil type A was coarser than B
and C and in order to achieve a homogeneous medium
for this soil, other conditioning agent such as polymer and bentonite was needed. Concentration (Cf) of
foam agent for all samples is 3%. Index f at curves
indicates foamed sands. The lowermost curve is for
unconditioned sand, initially prepared in a loose state.
At low stress level the void ratio lies just below
the maximum void ratio for the sand. As the stress
increases, the void ratio reduces by a small increment,
but the sand remains in a loose to medium density.
(Demonstrating the static stress is not sufficient for
sand compaction). The upper curves are for different
tests on foam/sand mixtures. The differences between
the initial void ratios at low stress levels are entirely
due to the amounts of added foam.At last stage the void
ratio of the sand/foam mixture is still well above the
void ratio at 0% relative density. It was unsurprising
that sand/foam mixtures could be made at high void
ratios, but it was quite unexpected that such high void
ratios could be sustained at a remarkably high stress
level. Note that the sand/foam mixture has a truly composite action rather than the sum of the component
parts. Sand would have been compacted to a much
lower density, and the foam by itself would have been
crushed at such a stress but the sand/foam mixture
is stable in a remarkably loose state. This may have
fundamental implications for tunneling operations.
According to performed tests, the foam/sand mixtures void ratio are greater than maximum void ratio
of dry sand and this characteristics of foam/sand mixtures is independent with respect to the foam types.
With increasing of FIR the void ratio increases intensively but increasing of FER would have negligible
effects on void ratio. In figures 6, 7 two curves show
different behavior with respect to other curves. These
two curves are Bf22 and Cf10 tests that are prepared by
foam C. The other samples were prepared with foam D.
Void ratio in foamed sands has been composed by

two components: air and liquid void ratio, majority of
void ratio in foam/sand mixture consists air void ratio.
If the foam bubbles are stable, the compressibility of
bubbles could lead to high compressibility of mixture.
This amount of compressibility via water drainage will
be not possible and also it need more time to drain. The
role of consolidation settlement is much lesser than
immediate settlement (Fig. 5).
According to figures 6, 7, the samples that are prepared by foam D show lower compressibility and lesser
sensitivity with respect to samples that are prepared by
foam C.
This phenomenon presumably is related to low stability of foam D against loading. But different drainage
times for two types of agents are not considerable. To
investigate this phenomenon, the volume change of
this mixture was verified by each volume change component (air and water). In figure 8, the void ratio
changes for soil and its component, air and liquid, have
been drawn with respect to pressure for two sand (B, C)
which mixed by foam type C and D. The initial and
final values are used for this plot. General tendency of
void ratio changes is distinct. It is clear that the water
void ratio variations for samples which prepared with
foam D (Figs 8a, 8c) are grater than air void ratio variations. The increasing of air void ratio has being seen
in some cases (Bf 20) because of excess sample drying
in drainage process. Although in Bf 22 and Cf 10 tests
(Figs 8b, 8d) air void ratio decrease more than water
void ratio. It means that the amount of drained water is
lesser than previous cases, and the majority of volume
changes were taken place via compressibility of foam
bubbles.
The air void ratio changes are compared in figures 9,
10. The air void ratio for samples with foam D gives
either a negligible change or a small increase (except
Cf 7). It can be concluded that the foam D increase
the water content and consequently the void ratio of
mixture preserving the homogeneity. But the compressibilitys is lesser because of low stability against
loading and during the test process foam bubbles wash
304
Figure 9. Air void ratio versus vertical stress for sand B.
Figure 10. Air void ratio versus vertical stress for sand C.
out via water drainage in second stages of tests. Foam

C also cause the increasing of water content in the
mixture, moreover the sample is high compressible
consequently shows a well sensitivity with respect to
loading (Figs 6, 7). This difference between two types
of foam agents are related to their chemical compositions. Note that the similar shear strength was
obtained for samples prepared by two different agents
(Sabetamal 2006). But depends on their compressibility different settlements were obtained. Thus, the foam
D must be improved for thats low stability time and
low compressibility in order to use in TBM machine
in sandy soil.
6
Figure 8. Void ratios variation versus vertical stress for sand

B,C with two different foam agents.
CONCLUSION
1. Index tests on various foam agents show that the

main differences between these agents are on stability or drainage time. The polymer increase the
stability time significantly.
2. One of the foam/sand mixtures characteristics is
high compressibility. Existence of foam bubbles in
the soil skeleton increase void ratios, this feature for
all foam agents is common but the behavior of sample during loading is affected by foam type. Thus,
use of same foam agent for different condition is
not suitable.
305
3. Large part of void ratio in foam/sand mixture consists of air void ratio. If the foam bubbles are stable
against vertical stress and during the test process
foam bubbles do not wash out via water drainage,
the compressibility of bubbles could lead to high
compressibility of mixture.
4. Some foam agents such as foam D increase the void
ratio at first but because of their unfitted chemical compositions they can not sustained against
loading.
5. With increasing of soil particle size at uniform gradation, use of foam alone as a conditioning agent
is not sufficient as the foam bubbles easily escaped
from the soil voids and assembled on soil mass.
Consequently the composite action of soil/foam
fails. So in coarser soils other conditioning agent
such as bentonite as a filler of voids must also be
used.
REFERENCES
Defence Standard 4240. 2002. Foam Liquids,Fire Extinguishing (Concentrates, Foam, Fire Extinguishing). Ministry of Defence. Issue 2. USA.
EFNARC, 2005. Specification and guidelines for the use of
specialist products for soft ground tunneling.
Head, K.H. 1994. Manual of soil laboratory testing: shear
strength, Compressibility and Permeability. 2nd edition,
Vol.2, Pentech press: London.
Milligan, G.W.E. 2000. Soil conditioning and lubrication in
tunneling, pipe jacking and micro tunneling. A state of art
review. web site: http/www.civils.eng.ox.ac.uk/research/
pipejacking.htm
Psomas, S. 2001. Properties of foam/sand mixtures for tunneling applications: M.Sc. thesis, St. Hughs college,
Oxford University. London: England.
Sabetamal, H. 2006. Study of soil conditioning on granular
soils for EPB Tunneling: M.Sc. thesis, Faculty of Civl
engineering, University of Tabriz. Tabriz: Iran.
306
Study on earth pressure acting upon shield tunnel lining in clayey and sandy
grounds based on field monitoring
T. Hashimoto, G.L. Ye, J. Nagaya & T. Konda
X.F. Ma
ABSTRACT: In this study, firstly, a series of field monitoring data on earth pressure in soft clay, hard clay
and sand ground are analyzed respectively. In the soft clay ground, the earth pressure fluctuated with backfill
grouting pressure at first, it settled down toward a steady value between (static vertical pressure Pv0 cohesion
C) and distributed uniformly over the ring finally. In the hard clay ground, the earth pressure was more greatly
influenced by the backfill grouting, especially for the lateral earth pressure in hard clay with 2C/pv0 0.5. In
the sand ground, although the earth pressure was also influenced by the backfill grouting, the distribution was
relatively uniform than in hard clay ground since the hydraulic pressure accounts for a large portion in earth
pressure. The insights obtained from this study can contribute to an improvement of load considerations in shield
lining design.
INTRODUCTION
In the design code of shield tunnel in Japan (JSCE,

1996), the earth pressure acting upon the segment
lining is calculated by the overburden pressure or
Terzaghis loosening earth pressure according to the
stratum condition and the overburden height only.
However, it is known that the earth pressure is also
influenced by the construction conditions (e.g. backfill grouting, position adjusting of shield machine),
and the interaction between the ground and man-made
structures (e.g. tunnel lining, pile foundation). In most
cases, these factors work together and undistinguishable. Therefore, nowadays the mechanical behavior of
the earth pressure upon shield tunnel lining has not
been clearly clarified yet.
Some researches have been done on this problem in
the last decades. After the Terzaghis theory on loosening earth pressure, Murayama (1968) studied the
vertical earth pressure in sandy layers by trapdoor tests.
According to the test finding that the sliding surface
is similar to logarithm spirals he proposed a formula
to calculate the vertical earth pressure. However, the
up-to-data shield technology equipped with precise
pressure control system at cutter face and simultaneous backfill grouting system makes it possible to build
a tunnel without loosening the surrounding ground.
Therefore, the actual earth pressure cannot be correctly predicted by conventional methods (Ohta et al.,
1997, Hashimoto et al., 1997). Moreover, Suzuki et al.
(1996) reported that the maximum loads occurred during backfill grouting in a shield tunnel with large
overburden.
In this study, firstly, three cases of field monitoring on earth pressure in soft clay, hard clay and sand
grounds are analyzed carefully. And focus is set on
the long-term behavior. The results distributions of
earth pressure, axial force and bending moment of
lining are compared with the calculated value by
conventional design method respectively. And more
than twenty measurement data are summarized and
organized according to the strength of the ground to
find out some empirical rules of earth pressure. The
insights obtained from this study can contribute to
an improvement of load consideration in shield lining
design.
2
BRIEF DESCRIPTION OF MONITORING

SITES
In order to clarify the characteristics of earth pressures

acting upon the linings in clayey and sandy grounds,
three typical monitoring jobs are chosen from three
307
Table 1.
Descriptions of shields and geology conditions.
Site name
Shield type
Kadoma
(Soft clay)
Osakajo-A
(Hard clay)
Osakajo-B
(Sand)
Mud-soil
pressure
balanced
Earth pressure Earth pressure

balanced
balanced
Segment type Ductile

Shield
5300 mm
diameter
RC
5300 mm
RC
5300 mm
Backfill
grouting type
Simultaneous
Simultaneous
Simultaneous
Overburden
height
14.09 m
28.2 m
16.8 m
Around soil
type
Alluvial clay
Diluvium clay Sand
SPT-N value
Unconfined
compressive
strength
35
89
170 200 kPa 540 kPa
>50
types of ground respectively, namely, soft clay, hard

clay and sand grounds. Some basic information on the
shields and geological conditions are listed in Table 1,
and the soil profiles of monitoring sites are shown in
Figure 1. Moreover, two comparing monitoring sections were setup at Osakajo-A and Osakjo-B sites to
check the influence of backfill grouting. In order to
obtain a reliable earth pressure data, a pad type earth
pressure cell (Hashimoto et al., 1993), as shown in
Figure 2, was adopted in all monitoring jobs. The
water pressures were recorded by piezometers from
the grouting holes.
3
FIELD MONITORINGS OF EARTH

PRESSURES ACTING UPON LINING
The time histories of observed earth pressures around

the lining in 3 types of ground are shown in Figure
3 5. The left part of the figures represents the shortterm data, and the right part is the long-term one. The
earth pressure and water pressure are noted as EP and
WP hereinafter. Without mentions, the earth pressure
means total earth pressure.
3.1
Figure 1. Soil profiles at monitoring sites.
In soft clay ground
Figure 3 shows the changes of observed EPs and WPs

at crown, left spring-line, right spring-line and invert
of the lining in soft clay ground. After the tail passing, the EPs fluctuated greatly by backfill grouting
pressure, and the fluctuation almost disappeared at
7th rings after tail passing, which indicated the extent
of influence from grouting holes is about 7 rings in
such a ground condition. After that, the EPs and WPs
decreased gradually in the first 1 2 months, and then
Figure 2. Pad type earth pressure cell.
the WPs remained unchanged while the EPs turned

to increased little by little, finally reached a constant
status (except R-spring). These final constant values
lie between (static pressure P0 cohesion C). Above
phenomena can be explained as following. The excess
porewater pressure adjacent to lining, which was generated during shield advancing and backfill grouting,
dissipated in the first 12 months resulting dropdowns
of EPs and WPs, and then the excess porewater pressure in the farer surrounding soil dissipated along with
the decoration of soil skeleton, resulted in a very slow
buildup of effective EPs upon lining.
308
Table 2.
Backfill grouting of Osaka-A shield
Injection pressure
Injection ratio
Grouting material
Case 1
Case 2
150 kPa
139%
Standard strength*
50 kPa
100%
Low strength
* Two components grouting material with a gel time less than

10 seconds.
Table 3.
Figure 3. Observed earth pressure acting upon lining in soft

clay ground (Kadoma Shield).
Backfill grouting of Osaka-B shield
Injection pressure
Injection ratio
Grouting material
Case 1
Case 2
300 kPa
135%
Standard strength
170 kPa
100%
Low strength
Figure 4. Observed earth pressure acting upon lining in hard

clay ground (Osakajo-A Shield).
3.2
In hard clay ground
Figure 4 shows the changes of observed EPs and WPs

at two compared cases in hard clay ground. In Case 1, a
regular backfill grouting was implemented. In Case 2,
in order to eliminate the influences (the squeezing
force to the ground) of backfill grouting, a special
grouting with smaller injection pressure and injection
ratio was used, as shown in Table 2. Since the longterm changes in both cases were influenced by the
parallel shield passing, earthquake and concrete casting of invert and arcade, and showed increasing trends,
the attention was paid to the changes before the 2nd
shield passing. In Case 1, after all the EPs primarily
climbing up to 300 400 kPa under the influence of
backfill grouting, the EPs at crown and invert turned
to decrease. On the other hand, the EPs in Case 2
was smaller (about 200 kPa) at first and then turned
to be increased by backfill grouting of next several
Figure 5. Observed earth pressure acting upon lining in

sandy ground (Osakajo-B Shield).
rings. Comparisons between two cases show that the

final differences were very small. The final EPs at
the crown in both cases were 370 380 kPa. The figure shows that the main part of each EP at different
location was determined at the first 10 rings, which
indicates that the EP largely depends on the backfill
grouting in the hard clay ground. And it also can be
seen that the settling down of EP only needed several
days, much quicker than in soft clay.
3.3
In sand ground
The same as that in Osaka-A shield, here also two

cases of comparing measurements were carried out,
as shown in Table 3. Figure 5 shows the changes
309
Figure 7. pv /pv0 vs. 2C/pv0 and ph /pv vs. 2C/pv0 (clayey
ground).
Figure 6. Conventional design method in Japan (effective

stress method).
of observed EPs and WPs in both cases. In Case 1,

right after the backfill grouting of current ring the
EPs was 100 150 kPa, then rose up to 200 kPa after
the grouting of next 2 3 rings, and further increased
to 200 250 kPa 4 months later. On the other hand,
in Case 2, right after the backfill grouting of current ring the EPs were only 80 110 kPa, and did not
increased significantly after next several rings. The
final values were 120 160 kPa even 4 months later.
The figure also shows that although the long-term EP
in both cases influenced by the parallel shield passing
and earthquake, and therefore showed an increasing
trend, however, the magnitudes did not change a lot
from those after the first several rings backfill grouting. And comparing to the EPs in clayey grounds, the
settling down of EPs in the sandy ground was much
faster.
shown in APPENDIX. The EPs over the ring in all

ground conditions were relative uniform over the ring
by comparing with the design values. Consequently
the axial force and the bending moment were smaller
than the design ones. As mentioned in previous session, the EP was greatly influenced by the backfill
ground in relative hard ground, and it is not difficult to
image that the backfill grouting can flow into the tail
void quickly and cover the whole ring with a uniform
pressure. Therefore, the uniform distribution of EPs is
directly related with the backfill grouting.
5
RELATION BETWEEN EARTH PRESSURE

AND GROUND CONDITION AND ITS
INTERPRETATION
Besides aforementioned three measurement data, we

also collected more than twenty data (long-term data)
from other monitoring jobs in Japan, including the
shield tunnels for subway, sewer, power-supply and
communication, and summarized them according to
the strength of the ground. The authors would like to
emphasize that based on these limited data it is still
difficult to execute a quantitative analysis. Therefore
in this session a qualitative analysis on the behavior of
EPs in various ground conditions is given out.
5.1 In clayey ground
COMPARISON BETWEEN DESIGN AND

OBSERVATION (HASHIMOTO ET AL. 2002)
In Japan, the conventional model and the bedded frame

model are main design method for shield lining. In
most common situation, such as the original designs
of the three tunnels in previous session, the engineer prefers the conventional model for the sake of
convenience. We also used conventional model for
comparison in this study. Figure 6 is a conceptual figure of conventional model, which considers the EP and
WP as well as soil reaction as lining loads.
The comparisons between the observed EP and
member forces as well as the calculated values are
The relation between pv /pv0 and the normalized

strength 2C/pv0 is plotted in Figure 7(a), where pv is
the effective vertical EP at the crown, pv0 is the effective overburden pressure, pv0 is the total overburden
pressure, and C is the cohesion. It is well known that
in natural clayey ground, there has such an empirical
relation as qu (=2C) = 0.3 0.4pv0 . When the ground
is unconsolidated with a very small 2C/pv0 (<0.3), the
tail void is easy to collapse immediately after tail passing, and cannot be easily filled by the backfill grouting
in time. Therefore, the surrounding soils will yield and
most of the overburden weight will act upon the tunnel
(pv /pv0 1). And when the ground is consolidated or
310
a little over-consolidated (2C/pv0 0.5), the grouting

material can fill the tail void in time before collapse.
On the other hand, when the stiffness of the grouting
material and surrounding ground are close, their interaction becomes active. Consequently, pv /pv0 depends
on the grouting pressure and varies largely according
to the construction conditions. Furthermore, when the
ground is stiff enough (2C/pv0 becomes large), the
shrinkage of grouting material during hardening will
be larger than the deformation of surrounding ground.
The backfill grouting has minor effect on the EP, resulting in a small pv /pv0 . The data in Figure 7(a) indicate
those kinds of phenomenon. For a better understanding, two boundary lines are drawn in the figure, and
for the areas lacking of data broken lines are drawn.
The relation between ph /pv and the normalized
strength 2C/pv0 is shown in Figure 7(b), where ph is
the effective horizontal EP at the spring line, and pv
is the effective vertical EP at the crown. It is found
that when 2C/pv0 is small (<0.3), ph /pv = 0.45 0.8.
Considering the coefficient of lateral earth pressure at
rest (K0 ) of clayey ground is about 0.5 and the location of spring line is deeper than the crown, such values
of ph /pv are rational. When 2C/pv0 0.5, the value
of ph /pv scatters in a wide range, implying that the
circumferential distribution of earth pressure in such
ground also depends on the backfill grouting and other
construction conditions.
5.2
In sandy ground
The relation between pv /pv0 and the equivalent SPT

(Standard Penetration Test) N value which is
regarded as a strength index is plotted in Figure 8(a).
By applying the Terzaghis loosening earth pressure
to an assumptive shield tunnel, we can get a curve of
loosening earth pressure against SPT-N value, as the
dot line in Figure 8(a). However, due to the influences
in construction process, the EP acting on the lining
will vary above or below the theoretical line. When the
ground condition is relatively good and the loosening
earth pressure is small, a proper backfill grouting can
reduce the EP dramatically, otherwise a large grouting pressure may remain on the lining. The former
phenomenon often occurs in the sand ground mingled
with cohesive silt/clay. The data with SPT-N between
4090 in Figure 8(a) are just some vivid instances.
Furthermore, when the ground becomes very dense
sand or gravel, similar to the clayey ground with large
2C/pv0 , the shrinkage of grouting material may larger
than the deformation of surrounding ground, and the
EP consequently becomes very small, such as the date
with SPT-N > 100 in the figure.
Although there is no enough observed data in the
SPT-N < 40 ground, we can make an estimation from
the knowledge of soil mechanics. When the soil is very
loose, the dependency on construction shall be weakened (the same as the soft clay ground), and then the
Figure 8. pv /pv0 vs. SPT-N and ph /pv vs. SPT-N (sandy
ground).
distribution of EP will get close to the theoretical loosening earth pressure, as the broken line in Figure 8(a).
For a better understanding, the authors also draw two
boundary lines in the figure.
The relation between ph /pv and the equivalent SPTN value is shown in Figure 8(b). It is found that
ph /pv scatters between 0 0.6, indicating that the circumferential distribution of earth pressure in sandy
ground also depends on the backfill grouting and other
construction conditions.
The shapes of the boundary lines in Figure 7(a) and
8(a) have the same characteristics: two ends are narrow
while the middle is wide. This kind of shape clearly
tells us that when calculating the EP in the ground with
medium stiffness, more attention should be paid to
the backfill grouting and other construction conditions
than those in soft or very hard ground. It should be
pointed out that, however, the backfill grouting must
be considered when dealing with the settlement in soft
clay ground (Hashimoto et al. 1999).
6
CONCLUSIONS
In order to clarify the mechanical behavior of earth

pressure acting upon shield tunnel lining in various
ground conditions, saying soft clay, hard clay and sand
grounds, a series of field monitoring has been carried
out. By analyzing the observations carefully, following
conclusions were obtained.
1. The earth pressure is influenced by the injection of
backfill grouting at 7 8 rings away in the case of
simultaneous backfill grouting.
2. In the soft clay ground, the earth pressure fluctuates
with the backfill grouting in the early phase. However, it settles down to a steady value between (static
pressure Pv0 cohesion C) finally, regardless of
the backfill grouting pressure.
3. In the hard clay and sand ground, the initial earth
pressure that builds up gradually by backfill grouting remains in the long-term earth pressure. In other
words, the earth pressure depends on the backfill grouting to such an extent that sometimes the
311
earth pressure will be larger than the prediction by

Terzaghis loosening earth pressure.
4. The settling down of earth pressure in soft clay
ground needs several months, while those in hard
clay and sand grounds only need several days or
even several hours.
5. The distributions of earth pressure are more uniform than predictions by conventional design
method in all ground conditions. In other word,
the bending moment is apt to be overestimated,
especially in the sandy ground where hydrostatic
pressure plays a dominant role in earth pressure.
Therefore, it is suggested that when designing the
lining, the influence of backfill grouting should be
taken into consideration.
6. By analyzing more than twenty monitoring data, it
is found that in the clayey ground, if 2C/qu < 0.3,
a large portion of the overburden will act upon
the lining. If 2C/pv0 0.5, the magnitude and distribution of earth pressure depend largely on the
backfill grouting. And in the sandy ground, if the
equivalent SPT-N value lies between 40 80, the
magnitude and distribution of earth pressure also
depend largely on the backfill grouting.
Ohta, H., Shiotani, T., Sugihara, K., Hashimoto, T. &

Nagaya, J. 1997. Consideration of Design Earth Pressure
for Shield Tunnel Based on Measurement. Proc. of Tunnel
Engineering, JSCE. Vol.9. pp.3742. (in Japanese)
Suzuki, M., Kamada, T., Nakagawa, H., Hashimoto, T. & Satsukawa,Y. 1996. Measurement of earth and water pressure
acting on the great depth shield tunnel segments. Geotechnical aspects of underground construction in soft ground,
IS-London, Mair & Taylor (eds), Balkema, pp. 613619.
APPENDIX
REFERENCES
Hashimoto, T., Nagaya, J. & Konda, T. 1999. Prediction of
ground deformation due to shield excavation in clayey
soils. Soils and Foundations, Vol.39, No.3, pp.5361.
Hashimoto, T., Nagaya, J., Konda, T. & Tamura, T. 2002.
Observation of lining pressure due to shield tunneling.
Geotechnical aspects of underground construction in soft
ground, IS-Toulouse, Kanster et al. (eds), Specifique,
pp.119124.
Hashimoto, T., Yabe, K., Yamane, S. & Ito, H. 1993. Development of Pad type earth pressure cell for shield segment.
Proc. of 28th annul report of JGS, pp.20552058. (in
Japanese)
Japanese Society of Civil Engineers (JSCE). 1996. Japanese
Standard for Shield Tunneling. 3rd edition.
Murayama, S. 1968. Earth pressure on vertically yielding
section in sand layer. DPRI Annuals, No.11(B), Kyoto
University, pp.549565.
Figure A1. Comparison of observed and designed values of

earth pressure and member forces (Hashimoto et al, 2002).
312
The double-o-tube shield tunnel in Shanghai soil

C. He
Department of Geotechnical Engineering, School of Civil Engineering, Tongji University, Shanghai, P.R. China
L. Teng
Shanghai Tunnel Engineering CO. LTD, Shanghai, P.R. China
J.Y. Yan
Shanghai Metro Operation CO. LTD, Shanghai, P.R. China
ABSTRACT: The high building density of shanghai and the wish to construct bigger city transport system lead
to an obvious conflict. The double-o-tube shield tunneling was introduced to save underground space. In 2002,
Shanghai has built its first double-o-tube shield tunnel in No.8 line. The stress and displacement distribution
around the double-o-tube shield tunneling was investigated from an in-situ test.
GENERAL INSTRUCTIONS
Shanghai, as one of the biggest city in China, the

density of building is quite high. But with the big
development of urban city, more and more building
and transportation constructions are needed. So the
questions that use limited space of shanghai to satisfy
all kinds of needs come out these years. In subway
area, the double-o-tube shield tunneling was introduced to save underground space. In 2002, Shanghai
has built its first double-o-tube shield tunnel in No.8
line. Figure 1 shows the double circular shield which
Figure 1. Break through of a double circular shield.
just finished one part of tunnel and broke through the

working shaft.
The origin of the double circular shield tunnel may
be traced to 1981, when a basic patent was applied
for in Japan. The patent was registered in 1987, and a
horizontal double circular shield tunnel field trial was
performed in the same year. In 1988, a vertical double
circular shield tunnel trial was conducted in the field.
In the construction of the national road No. 54 tunnel in Hiromiosa, a double circular shield, which was
designed and manufactured by Ishikawajima-Harima
Heavy Industries Co. Ltd, was used to build the first
double circular shield tunnel in the world with length
of 853.8 m (Moriya, Y., 2000).
The shield used in Shanghai subway was made in
Japan by Ishikawajima-Harima Heavy Industries Co.
Ltd and assembled in Shanghai by Shanghai TBM
Company. The width of shield is 11.12 m; the diameter
of a circular is 6.52 m. The diameter of mostly single
circular shields used in Shanghai is 6.4 m. The length
of shield is 12.76 m. The distance between the lining
and shield is 11 cm. Figure 2 shows the main dimension
of double-o-tube shield tunnel shield machine.
The double-o-tube shield tunneling Method is
applied for an earth pressure balanced shield machine
with interlocking spoke-equipped multiple cutters that
are positioned in the same plane. Adjacent cutters
rotate in the opposite directions to avoid touching or
smashing one another and are thus controlled synchronously. The double circular shield machine is
313
Figure 2. The dimension of the double circular shield.
equipped with cantilever-arm-type erector to erect

joint and panel segments, so it provides wide working
space. The double-o-tube shield tunnel is composed
of 11 prefabricate concrete segments (including the
columns in the middle of the tunnel) per section.
COMPARISON BETWEEN SINGLE AND

DOUBLE-O-TUBE SHIELD TUNNEL
Compared with the single circular shield, the most

important advantage of the double-o-tube shield tunnel is saving space. Most of Shanghai metro was built
under the road, and the space is limited when the
road is relatively narrow. Apparently, the double circular shield tunnel may pass narrower underground
corridors, and the impact on nearby structures is minimized (Moriya, Y., 2000 and R.C. Sterling, 1992).
Meanwhile, the double circular shield tunnel has an
optimized cross-section with a minimized section area,
enabling the most efficient use of underground space.
Moreover, the cross-passage between two circular
shield tunnels is unnecessary for a double circular
shield, and construction risk is therefore eliminated.
Figure 3 shows you the concept that double circular
shield tunnel can save space comparing with two single
tunnels.
On the other hand, the double-o-tube shield tunnel has its own disadvantages. The first one is cost.
China can produce the single circular shield now, but
the double circular shield is made in Japan and assembled in China. Moreover, the construction cost of the
double-o-tube shield tunneling is more expensive than
two single circular tunnels. Secondly, the double circular shield is not easy to be controlled in small curvature
route. Sometimes it needs extra load to balance the
shield. Thirdly, the speed of double-o-tube shield tunneling is slower than two single circular tunneling
which shared the same shield. The main reason is that
the linings fix spends more time.
Figure 3. Comparison between single shield tunnel and

double circular shield tunnel.
During the tunneling process, there are three important construction parameters: face pressure, speed
and grout volume. The face pressure which the
double-o-tube shield tunneling used is little higher
than which used in the single circular tunneling.
Higher face pressure can reduce the settlement while
tunnel is finished, however it would induce the higher
heave before the opening face.
The speed of the double-o-tube shield tunneling is
much lower than the single circular tunneling. The
main reason is the lining fix cost more time in the
double-o-tube shield tunneling. And maintaining a
certain low speed is efficient way for the settlement
control. According to the construction experience of
Shanghai, lower speed correspond smaller ground
settlement especially for shallow cover depth case.
Although grout pressure is more frequently mentioned in theoretical and numerical calculation of the
tunneling as the mechanical parameter, in the practical tunneling work of Shanghai, grout volume used to
present the effect of grout instead of grout pressure.
The main reason is that the grout pressure is unstable
and hard to control for real work. The grout volume
which the double-o-tube shield tunneling need is much
more than the volume which single circular tunneling
314
Figure 5. Gauges layout in cross section A.

Figure 4. Typical profile of ground condition in Shanghai.
need. But the grout volume is not a fixed value. It is

depend on different soil types, cover depth, grouting
point, and grout material. Even while all the factors
are same, the grout volume would still be different for
different company which in charge of tunneling. It is
more based on the experience but not theoretical result.
3
Along the tunneling route, most of the cover depth

varies from 6 m to 16 m. There are two typical cases
according to the different cover depth. One is 8 m cover
depth representing shallow cover. Another one is 16 m
for deep cover. Surface settlement due to tunneling is
significant different between these two cases.
GEOLOGY CONDITION
Since Shanghai is located near the confluent of

Huangpu River and Chang jiang (Yangzi River),
the soil has been progressively constituted by sedimentation and ended in a superposition of layers.
The Shanghai soil can be assumed as homogeneous
in most Shanghai land except partial place where
some soil layers are missed. In Shanghai area, the
mucky soil stratum is down to about 30 m deep
from ground surface. It is basically saturated fluidplastic or soft-plastic clay with low shear strength
(0.005 0.01 Mpa), high water content (above 40%),
high compressibility (0.5 1.0 MPa1), sensitivity
varying in 4 5, and evident rheological behavior
(Wang Zhen Xin & Bai Yun, 2004). The water table
is ranged from 0.3 m 1 m. In this very soft ground,
deep excavations for constructing underground metro
station and high building basement encounter many
difficulties in environment protection. The general
ground condition of Shanghai Metro Line is shown
in Figure 4.
There are mainly two soil types which
double-o-tube shield tunnel cross through. One is the
Grey muddy clay and Grey clay which are the typical soft clay layers. Another situation is in the place
where the Grey muddy clay is missed and Grey clay
is very thin. It means that layer Grey muddy silty
clay almost connects with layer Grey silty clay. So the
tunnel mainly crosses the silty clay.
IN-SITU TEST
4.1 Stress distribution in the soil due to the

double-o-tube shield tunneling
Geotechnical engineers discussed on a theoretical base
about the soil stress distribution due to double-o-tube
shield tunneling at the beginning of double circular
method applied in Shanghai. They try to compare the
soil stress distribution due to double-o-tube shield tunneling with the results from single circular tunneling.
Two theories came out. One opinion is that the soil
stress distribution from the double-o-tube shield tunneling could be assumed as the combination of soil
stress from two separate single circularities which their
location are coincident with double-o-tube shield tunnel. Other engineers prefer to believe that it should
be modeled from a bigger equivalent single circular
which share same center point with the double-o-tube
shield tunnel.
An in-situ test has been done by Shanghai Tunnel Engineering Co. Ltd to investigate the stress and
displacement distribution around the double-o-tube
shield tunneling. There are 4 sections in which the
displacement inspector and stress gauge were placed
along the line of tunneling.
Cross section A was designed for observing the soil
stress increment and water stress increment. Figure 5
shows the layout of gauges on cross section A.
Figure 6 shows the vertical soil stress increment in
the plane which in 1.5 m ahead of the opening face.
315
Figure 6. The vertical stress increment in 1.5 m ahead of the

opening face.
Figure 8. The settlement troughs with cover depth of 8 m.
Figure 7. The vertical stress increment distribution in 1.5 m

ahead of the opening face.
Figure 9. The settlement troughs with cover depth of 16 m.
There are three soil stress increment distribution line

in figure 7 with different depth. It is drawn from 4
inspectors at 3.02 m deep, 4 at 6.52 m and 6 at 10.78 m.
It is different tendency for these distribution lines. A
triangle shape which the largest soil stress increment
occurred on the middle vertical line of double-o-tube
shield tunnel is got for the top distribution line which
near to ground surface. For the bottom distribution line
which lies on the same depth level with tunnel horizontal central line, the stress increment in the middle
point of tunnel is smaller than the one in the center
point of two circles.
Figure 7 shows the soil stress distribution due to
the double-o-tube shield tunneling of one section. The
black line in figure represents the isoline soil stress
around the tunnel at the time while the opening face is
1.5 m behind.
From the result, the soil stress distribution induced
by tunneling is more closed to an ellipse than the combination of two single circularities or a big circularity.
And the effective influence width from double-o-tube

shield tunneling is mostly 2.5 times to width of tunnel.
4.2
Surface settlement due to the double-o-tube

shield tunneling
The requirement of surface displacement due to tunneling is +1 3 cm in Shanghai. In the beginning

of the double-o-tube shield tunneling, the settlement was surprisingly high. Many people suspected
this tunneling method at the beginning of doubleo-tube shield tunnel applying, however after some
projects; the engineers acquire the experience of the
double-o-tube shield tunneling, the settlement has
been under control in most conditions.
Figure 8 and Figure 9 give the surface settlement
trough due to tunneling in two cases. And each figure
shows two soil types which tunnel cross through.
316
Figure 11. The measurement results of building settlement

after tunnel crossing.
Figure 10. The plan view and transverse section of the

double-o-tube shield tunnel crossing building A.
It is hard to obtain equal volume loss for each circle

in the advance of tunneling, especially in the curve. So
from these two figures, it is quite normal phenomena
that the settlement trough is asymmetry to the center
of shield.
Concerning the influence of soil type, the maximum
settlement of silty clay is higher than the value of clay.
Meanwhile the settlement trough of clay is wider. The
main reason is that the influence area in the soft clay
is lager than the one in silty clay with the same construction conditions, not only in the cross section but
also in the longitudinal direction.
Comparing the data in different cover depth, the
maximum settlement of 16 m cover depth is less than
the one of 8 m cover depth. In the same time, the settlement trough of 16 m is wider than the one of 8 m
cover depth. This phenomenon is coinciding with the
Peck formula.
5
DOUBLE-O-TUBE SHIELD TUNNEL

CROSSING BUILDING
Here is an interesting project during the

double-o-tube shield tunneling. There are three key
construction phase of this double-o-tube shield tunnel
line. The shield need across three key section- pipe
channel, five-floor buildings and shallow cover depth
area. The current paper will represent the second key

section which tunnel crossing a five stories building.
Figure 10 shows the plane view of three key construction phases and the cross section of second one.
The distance of the building and the tunnel is very
close; it ranged from 1.2 m to 1.0 m. It is great risk
that the tunnel crosses a building with such close distance. Especially this building with raft foundation has
already had 0.3% inclination. According to Chinese
foundation code, the maximum inclination of building should be controlled under 0.4%. Meanwhile it
has been mentioned before that there are not enough
experiences of the double-o-tube shield tunneling
comparing with the single circular tunneling.
Shanghai Tunnel Engineering Co. Ltd take series
actions to control the ground settlement and reduce
the influence of tunneling to building raft foundation.
In the process of tunneling, the face pressure remains
high level. The grout volume happened in the tail is
larger than the value which used normally. The tunneling speed was maintained at very low in the whole
process. Moreover, extra grout was used while the tail
passed. With all these construction method, it results in
that the settlement and the inclination of the building
was controlled well.
There were 6 displacement inspectors installed in
the foundation of the building. The final measurement
data were given in Figure 11.
The blue number shows the location of displacement inspectors. The red number is the settlement
of raft foundation respectively corresponding to each
inspectors and the unit is centimeter. The measurement of settlement ranged from 0.32.2 mm. It means
that the tunneling is successful. The building is almost
intact and no change can be detected by eyes.
6
CONCLUSIONS
The double-o-tube shield tunneling has been applied

in Shanghai successfully. The settlement can be
317
controlled in a very low value. The construction is

more based on the experience but not theoretical result.
Furthering research need to be done to get a theoretical
guarantee for future work.
ACKNOWLEDGEMENTS
The work described in this paper was part of the
project funded by Shanghai Tunnel Engineering Co.
Ltd. The authors would like to acknowledge the collaboration throughout the project of Shanghai TBM
Company. Particular thanks are due to Tang xuan and
Wu huiming.
REFERENCES
Moriya, Y., 2000. Special shield tunneling methods in Japan.
In: Proceedings of the International Conference on Tunnels and Underground Structures, Singapore, pp. 24925.
Shanghai Tunnel Engineering Co. Ltd., 2006. The report of
double circular tunneling (1), Sept.
Sterling, R.C. 1992. Developments in excavation technology,
a comparison of Japan, the US and Europe, Tunneling and
Underground Space Technology. 7 (3), pp. 221235.
Wang, Z.X. & Bai, Y. 2004. Urban soft ground tunneling in
China experiences from Shanghai and other cities,
Proceedings of World Tunnel Congress and 13th ITA,
Singapore, pp. 2227.
318
Frozen soil properties for cross passage construction in

Shanghai Yangtze River Tunnel
X.D. Hu
A.R. Pi
Department of Civil Engineering, Anhui Institute of Architecture & Industry, Hefei, P.R. China
ABSTRACT: Eight cross passages between the two tubes of Shanghai Yangtze River Tunnel are constructed by
Artificial Ground Freezing Method. The formations around the tunnel are characterized with saline soil, which
raises such disadvantages as freezing-point depression, freezing phase lengthening, the growth rate decrease of
frozen bodies and the strength loss of the frozen ground. In order for successful construction, the properties of
the artificial frozen soils are made out by test, such as salinity, freezing point, uniaxial compressive strength,
thermal conductivity, frost heave ratio, thaw consolidation ratio and other related parameters. Based on the test
results some suggestions are given for freezing scheme design, the cross passages construction and freezing
process monitoring.
INTRODUCTION
As in many other tunnel projects (Crippa et al. 2006,

Hu et al. 2006 & Zhao et al. 2005), eight cross passages
between the two tubes of Shanghai Yangtze River Tunnel (see Figure 1) are constructed by artificial ground
freezing method. Shanghai Yangtze River Tunnel is
situated at the estuary of theYangtze River, where saltwater intrusion occurs frequently (Shen et al. 2003).
It is supposed that the saline concentration of soil at
the riverbed is higher than usual, thus the higher risk
in constructing the cross passages by artificial ground
freezing method. The properties of frozen saline soil

have a remarkable change (Roman et al. 2004) which
is disadvantageous to the construction. As a result, it is
of great necessity to proceed a research of the properties of the saline soil such as the salinity, the freezing
point of the saline soil and the frozen soil strength, etc.
The research concentrated on the basic geotechnical
parameters, the physical mechanic parameters and the
thermodynamic parameters of the designed soil layers,
including gray mucky clay ([4]), gray clay ([5]1), gray
clayey silt ([5]2), gray silty clay ([5]3) and gray clayey
silt ([7]1-1).
Figure 1. Layout of the cross passages of Shanghai Yangtze River Tunnel.
319
Table 1.
Results of geotechnical tests.
Soil layer
(undisturbed)
Water
Content
w (%)
Density
(g/cm3 )
Permeability
coefficient
K (cm/s)
Void ratio
e0
Plastic limit
wP (%)
Liquid limit
wI (%)
Salinity
()
Chlorinity
()
[4]
[5]1
[5]2
[5]3
[7]11
44.02
38.21
29.96
31.44
26.83
1.69
1.77
1.80
1.78
1.86
3.05 107
2.15 107
8.55 105
1.99 106
7.90 104
1.40
1.07
0.94
1.01
0.85
40.78
33.97
30.42
36.73
36.02
23.72
21.68
19.92
21.17
20.35
18.370
17.392
13.228
8.039
14.744
10.169
9.627
7.322
4.450
8.161
2 TESTS
Table 2.
2.1
Soil layer
8 C
15 C
20 C
25 C
[4]
[5]1
[5]2
[5]3
[7]11
3.14
3.19
4.18
3.07
4.29
4.07
4.82
5.26
3.95
5.44
5.29
5.47
6.17
5.16
6.55
5.94
6.58
7.07
6.60
7.61
Requirement and standard
The tests are proceeded according to the standards of

Peoples Republic of China as follows:
Standard for soil test method (GB/T50123-1999)
Coal Industrial Standard (MT/T593-1996)
2.2
Content
2.2.1 Basic geotechnical tests

The basic geotechnical tests include test of density,
water content, permeability coefficient, void ratio,
plastic limit, liquid limit, salinity and chlorinity.
Table 3. The relationship between uniaxial compressive

strength (MPa) and temperature T ( C).
2.2.2 Frozen soil tests

Frozen soil tests include:
1. Uniaxial compressive strength test
Uniaxial compressive strength tests are conducted
at the temperature of 8 C, 15 C, 20 C and
25 C respectively to achieve the uniaxial compressive strength, modulus of elasticity and Poissons ratio.
2. Frost heave test
Frost heave tests are conducted to achieve frost
heave ratio and frost heave force.
3. Thaw consolidation test
4. Thermal conductivity test
5. Freezing point test
6. Specific heat capacity test
3
3.1
Results of uniaxial compressive strength.
RESULTS
Basic geotechnical tests
Soil layer
Linear fitted relationship
[4]
[5]1
[5]2
[5]3
[7]11
= 0.1708 T + 1.706
= 0.1943 T + 1.7118
= 0.1704 T + 2.7725
= 0.2073 T + 1.1713
= 0.1965 T + 2.6328
R2 = 0.9844
R2 = 0.9901
R2 = 0.9982
R2 = 0.9641
R2 = 0.9949
temperature falls. Furthermore, these two variables

obey a favorable linear relationship. The linear fitted correlation between uniaxial compressive strength
and temperature T according to the test data is
presented in Table 3 and Figure 2.
(2) Modulus of elasticity
The result of the test indicates that, generally,
modulus of elasticity of the frozen soil rises when
temperature falls. Data acquired from the test is shown
in Table 4. The exponential correlation between modulus E of elasticity and temperature T could be fitted
in form of
The results of basic geotechnical tests are presented in

Table 1.
3.2
Frozen soil tests
3.2.1 Uniaxial compressive strength test

(1) Uniaxial compressive strength
Table 2 shows the result of uniaxial compression test, which indicates the fact that the uniaxial
compressive strength of the frozen soil rises when
as presented in Figure 3 and the coefficients a and b

are shown in Table 5.
(3) Poissons ratio
The result of Poissons ratio is shown in Table 6,
which demonstrates that when temperature falls, the
Poissons ratio of the frozen soil tends to decline,
though this trend is not remarkable.
320
450
[4]
[5]2
[5]3
5
[4]
400
[5]1
Modulus of elasticity /MPa
Strength /MPa
[7]1-1
4
3
[5]1
350
[5]2
300
[5]3
250
[7]1-1
200
150
100
50
2
0
-5
-10
-15
-20
-25
-30
-5
Temperature/C
-10
-15
-20
Temperature/C
-25
-30
Figure 2. The relationship between uniaxial compressive

strength and temperature.
Figure 3. The relationship between modulus of elasticity

and temperature.
Table 4.
Table 6.
Result of modulus of elasticity.

Modulus of elasticity/MPa
Result of Poissons ratio.

Poissons ratio
Soil layer
8 C
15 C
20 C
25 C
Soil layer
8 C
15 C
20 C
25 C
[4]
[5]1
[5]2
[5]3
[7]11
53.1
58.2
50.7
57.1
79.2
78.7
97.9
82.4
93.2
155.7
146.5
165.2
155.7
161.1
238.1
246.8
269.0
273.5
306.1
401.0
[4]
[5]1
[5]2
[5]3
[7]11
0.281
0.277
0.298
0.259
0.266
0.263
0.251
0.269
0.242
0.244
0.230
0.229
0.248
0.217
0.218
0.214
0.204
0.221
0.206
0.180
Table 5. The coefficients a and b in the relationship between

modulus E (MPa) of elasticity and temperature T ( C).
Table 7. The results of frost heave force and frost heave ratio
tests.
Exponential fitted relationship

2
Soil layer
[4]
[5]1
[5]2
[5]3
[7]11
23.173
26.975
20.923
23.848
37.176
0.0921
0.0907
0.1005
0.0985
0.0945
0.9726
0.9939
0.9824
0.9796
0.9989
3.2.2 Frost heave test

Frost heave ratio test: frost heave test without axial
confinement so that the specimen can expand freely.
The relationship between axial displacement and time
is measured at the time according to code requirement. The maximum of frost heave max should also
be recorded. Frost heave ratio of the specimen means
the ratio between the maximum of frost heave max and
initial length of the specimen.
Soil layer
Frost heave force

(MPa)
Frost heave ratio

(%)
[4]
[5]1
[5]2
[5]3
[7]11
0.66
0.78
0.84
0.71
0.69
7.93
6.84
7.26
6.35
6.48
Frost heave force test: frost heave test proceeded

with displacement confinement. Lengthwise confinement is applied to the upper end of the specimen and
then frost heave force is measured by load sensor.
The relationship between frost heave force and time is
recorded during the whole process to get the maximum
of frost heave force max .
The maximum of frost heave ratio and frost heave
force of each soil layer is presented in Table 7.
321
Table 8. The result of the thaw consolidation ratio.
Table 10. The result of freezing point.
Soil layer
Thaw consolidation ratio (%)
Soil layer
Freezing point ( C)
[4]
[5]1
[5]2
[5]3
[7]11
6.83
7.09
7.76
7.25
6.05
[4]
[5]1
[5]2
[5]3
[7]11
2.5
2.3
1.7
2.1
1.7
Table 9. The result of the thermal conductivity.
Table 11.
Result of undisturbed soil specific heat capacity.
The thermal conductivity (W/(mK))
Soil layer
Specific heat capacity/(gK)
Soil layer
positive
negative
[4]
[5]1
[5]2
[5]3
[7]11
1.39
1.43
1.47
1.42
1.54
1.62
1.75
1.68
1.74
1.81
[4]
[5]1
[5]2
[5]3
[7]11
1.56
1.62
1.69
1.65
1.51
3.2.3 Thaw consolidation test

The temperature of the specimen is 8 C, the hot
end temperature is 40 2 C. The dimension of the
specimen is
79.8 40 mm. The thaw consolidation
ratio is determined as follows:
where a0 = thaw consolidation ratio (%); h0 =

quantity of thaw consolidation (mm); and h0 = initial
height of the specimen (mm).
The result of thaw consolidation test of each soil
layer is shown in Table 8.
3.2.4 Thermal conductivity test
The result of the thermal conductivity is presented in
Table 9. In this table, positive is for the unfrozen soil
and negative for the frozen soil.
3.2.5 Freezing point test
Generally speaking, the crystallization from liquid to
solid of the water in the soil undergoes three stages:
First a small group of molecule is formed which is
called the crystallization center or growth point, then
it grow up to a bigger crumb called crystal nucleus,
which eventually develops to ice crystal. The temperature of the formation of the ice crystal is called freezing
point or ice point. The crystallization center is formed
at a certain temperature below the freezing point. As
a result, the formation of frozen soil is consisted of
four stages, i.e. supercooling, jump, invariableness and
degradation. At the stage of jump, the electric potential
will suddenly reduce, and then become stable at a certain number at the temperature when freezing begins.
According to this principle, the freezing point can be

calculated as follows:
where T = freezing point ( C); V = the stabilized value

after the jump of the hot electric potential ();
and K = the demarcation coefficient of thermoelectric
couple ( C/).
The result of the freezing point of each soil layer is
presented in table 10.
3.2.6 Specific heat capacity test
The specific heat capacity test is performed with the
specific heat capacity testing device DTBR-01. Introduce coolant of the same temperature and volume into
both sides of the equipment. Temperature sensors are
previously put into coolant.The specimen with temperature sensor embedded is put into one side. According
to the law of conservation of energy, the energy for one
degree changing can be determined by the relationship
between the three values of temperature. The result is
presented in table 11.
DISCUSSION AND SUGGESTIONS
4.1 Salinity and its influence

The salinity of pore water in the undisturbed soil ranges
from 8.039 18.370% (see Table 1). To find out the
salinity difference between the soils near the sea and
far from the sea, an additional test was made for a soil
layer the gray clayey silt ([7]11 ). In the result, the
salinity of the specimen of the layer in the urban area
(at the site of the restoration project of the collapsed
322
tunnels of Shanghai Metro Line 4) is 6.320%, while

that of the same layer at the estuary is 14.744% (see
Table 1). Therefore, a conclusion could be drawn that
the salinity of pore water in the strata near the sea is
higher than that distant form the sea.
Influence of salinity on frozen soil properties could
be freezing-point depression, freezing phase lengthening, the growth rate decrease of frozen bodies and the
strength loss of the frozen ground, on which a great
attention must be paid.
Given lack of enough test data, no clear relationships have been found between the salinity and frozen
soil strength, frost heave force, frost heave ratio, thaw
consolidation ratio and other parameters, as well as
freezing point, but further tests are being performed
by the authors.
4.2
Freezing-point depression
The freezing point ranges from 1.7 to 2.5 C (see

Table 10). It is found that the freezing-point depression
of the soil at the estuary is greater than that distant from
the sea. Fro example, the freezing-point temperatures
of the gray clayey silt [7]11 near the sea and distant
from the sea are 1.7 C (see Table 10) and 1.4 C
(Xiao, et al. 2003), respectively. On the other hand, soil
layers are sorted according to the descending order of
the magnitude of freezing-point depression, i.e. mucky
clay, clay, silty clay and clayey silt.
Freezing-point depression causes a smaller thickness of the soil frozen wall than the thickness when the
freezing point is considered as 0 C. Attention must be
paid upon the thickness loss of the frozen soil wall due
to freezing-point depression.
4.3
Frozen soil strength
The uniaxial compressive strength of the frozen soils

and the temperature of the specimen bear a favorable linear relationship, with regression coefficient all
higher than 0.96. Meanwhile, the moduli of elasticity
of the frozen soils have an exponential correlation with
the temperature, with an average regression coefficient
of 0.985. The regression formulas can be used to calculate the strength values at any temperature within
the range of the test temperature.
A suggestion here is that the regression formulas
should be used to calculate the bearing capacity and
deformation of the frozen wall.
4.4
Frost heave
great emphasis should be laid on the design and application of the construction. It is suggested to install
a steel reinforcement frame around the opening as a
protection device to prevent the tunnel lining from
excessive deformation due to the frost heave.
4.5 Thaw consolidation
The ratio of the thaw consolidation ranges from 6.05%
to 7.76%. The thaw settlement of the ground around
the cross passages could be remarkable. Therefore, it
is advised to perform enforced thawing as an effective
measure to reduce thaw settlement.
4.6 Thermal properties
The thermal conductivity for unfrozen soil ranges from
1.39 to 1.54 W/(mK), and 1.62 to 1.81 W/(mK) for
frozen soil. The soils are hard to be frozen, i.e. the
growth rate of the frozen body could decrease or the
freezing phase could be lengthened, because the coefficient of thermal conductivity is unfavorably low. It
is necessary to enhance freezing process monitoring
and to use the monitored data to judge the size of the
frozen wall instead of rude estimate of the freezing
phase.
5
CONCLUSIONS
Although the strength of the frozen soils at the estuary of the Yangtze River is fairly high, attention
must be paid upon the disadvantageous influence of
freezing-point depression due to the high salinity on
the calculation of the thickness of the frozen soil walls
for constructing the cross passages.
Further study should be carried out to find out
the influence of salinity of pore water in the saline
soils upon the geotechnical and physical-mechanical
properties discussed before.
ACKNOWLEDGEMENTS
The work were supported by the National Natural Science Foundation of China (No. 50578120),
the National High Technology Research and Development Program of China (863 Program) (No.
2006AA11Z118) and Shanghai Leading Academic
Discipline Project(Project Number: B308).
REFERENCES
The frost heave force ranges from 0.66 to 0.84 MPa.

The frost heave ratio ranges from 6.35 to 7.93%.
According to Code for Design of Soil Foundation of
Building in Frozen Soil Region (JGJ118-98), when the
frost heave ratio of the clay is higher than 6%, the soil
is categorized as highly frost-heaving soil. Therefore
Crippa C. & ManasseroV. 2006.Artificial Ground Freezing at

Sophiaspoortunnel (The Netherlands) Freezing parameters: Data acquisition and processing. GeoCongress
2006: Geotechnical Engineering in the Information Technology Age, Atlanta, Feb 26 Mar 1, 2006. Reston, VA:
ASCE.
323
Heilongjiang Province Academy of Cold Area Building

Research. 1999. Code for Design of Soil Foundation of
Building in Frozen Soil Region (JGJ118-98). Beijing:
China Architecture & Building Press.
Hu X.D. & Chen R. 2006. Construction technology of
freezing method applied to cross-passage of double-deck
cross-river road tunnel. Low Temperature architecture
technology. (5). 6466.
Roman L.T., Aksenov V.I. & Veretehina E.G. 2004. Characteristics of influence of the content salts on the frozen
soils strength. Journal of glaciology and geocryology
26(Suppl): 3538.
Shen H.T., Mao Z.C. & Zhu J.R. 2003. Saltwater Intrusion
in the Changjiang Estuary. Beijing: China Ocean Press.
The Ministry of Coal Industrial P.R.C. 1996. Coal industrial
standard: Physical-mechanical properties test of artificial
frozen soil (MT/T593-1996). Beijing.
The Ministry of Water Resources P.R.C. 1999. Standard

for soil test method (GB/T50123-1999). Beijing: China
Planning Press.
Xiao Z.H., Hu X.D., Pi A.R. & Liu R.F. 2003. Study on
uniaxial compressive strength of Shanghai soils under secondary freeze-thaw action. Rock and Soil Mechanics 27
(Suppl): 497500.
Zhao Y.H., Hu X.D., Zhao G.Q. 2005. Experimental study
on artificial frozen soil boundary detecting during crosspassage construction in tunnel. Progress in Electromagnetics Research Symposium, Hangzhou, China, 2226
August 2005. Cambridge:The ElectromagneticsAcademy.
324
The influence of engineering-geological conditions on construction of the

radioactive waste dump
J. Kuzma & L. Hrustinec
Slovak University of Technology, Faculty of Civil Engineering, Bratislava, Slovakia
ABSTRACT: Secure stability and reliable serviceability of radioactive dump is a difficult engineering problem.
Due to difficult geological formations determined mainly by great compressibility, low shear strength of soils,
and high ground water level, or great upward hydrostatic pressure will these demands increase. Influence of
required reliability and lifespan on the structure of these specific objects is considerable. We are trying to
contribute to a problem solving of these difficult and complicated problems in submitted contribution.
INSTRUCTION
The Republic Radioactive Waste Dump (RRWD)

located about 2 km from the area of the Mochovce
Nuclear Plant in south-western Slovakia, Central
Europe, was built to store waste with low and medium
activity arising from operation and disposal of nuclear
power facilities as well as from research institutes, laboratories, hospitals and other institutions engaged in
activities generating nuclear waste. Ensuring stability and reliable operation of RRWD in Mochovce is
a challenging engineering task. The dump is built in
demanding engineering-geological conditions determined in particular by high compressibility, low shearing strength of soil and high underground water level
and/or high lifting forces exerted in the basement.
Also required reliability and lifetime have a significant influence on layout of these specific structures.
This paper discusses a solution for this complex
engineering task.
The topic of how to secure a reliable storage of final
processed low and medium radioactive waste in the
long term is a complex interdisciplinary issue. The
time aspect is a significant factor affecting a complex
approach to this problem and specific periods of active
operation and planned lifetime of the repository must
be taken into account. The repository is built in fairly
difficult geological and hydro-geological conditions.
It consists of a complex of buildings and technological
facilities. In the first construction stage, two double
lines of storage boxes with the total surface area of
approx. 112.000 m2 were built. The general situation
around the dump and its adjacent territory is shown on
Figure 1. Section of the RRWD is shown on Figure 2.
Figure 1. View of placement of the Republic Radioactive

Waste Dump (RRWD) in Mochovce.
Figure 2. Section of the RRWD in Mochovce.
Such large structural work challenging in terms of

operational safety will have a significant impact on the
surrounding natural environment during its lifetime.
Extensive construction activities, active operation of
the repository and the final coverage will have a significant influence on a change in the balanced state
325
of the rock environment. This change in the balanced

state will cause deformation symptoms at the affected
territory with a number of structural buildings and
functional (technological) units. Knowing the extent
of the modified basement and construction at the territory concerned will be decisive to determine stability
of the territory and functionality of the already built or
planned drainage systems, engineering and technological distribution lines. This paper will discuss in detail
the issue of repository reliability in terms of assessing the basement according to requirements of group
I and II limit states, i.e. according to safety and usability of the structural work. A set of geotechnical issues
is related to the draft construction project of the final
repository coverage with its primary function to prevent water infiltration to premises housing the storage
boxes after the active operation of the structural work
until its lifetime.
Figure 3. Typical cross section of the 1st and 2nd double-line of storage boxes in RRWD in Mochovce.
modified by constructional activities. In terms of

assessing the foundation conditions, the engineeringgeological and hydrogeological conditions at the area
of interest can be assessed as difficult.
3
2
MORPHOLOGICAL, GEOLOGICAL AND

HYDROGEOLOGICAL CONDITIONS AT THE
REPUBLIC DUMP
Republic Radioactive Waste Dump is located in southwestern Slovakia, Central Europe. The terrain is
slightly folded with an unstable superficial water river.
The basement is made of sandy limestone clay with
spots of dusty fine-grain sand and clay silt. These spots
form lenses separated from each other. Sediments of
Ivanian strata (lower Panonium) with thickness of
11.523.0 m were identified at the northern part of the
repository. They are represented by sandy clay with
thin spots of dusty sands and isolated lenses of gravel
sands. Quaternary deposits are represented primarily
by deluvial rain-wash sand clay with thickness up to
3.0 to 4.0 m. The basement directly under the first
double-line of boxes is made of residual quaternary
deposits (clays), especially in the middle of the box
structure and under the bottom doubleline. Panonian
deposits (clays, sands) are found underneath.
Hydrogeological conditions in the RRWD basement
and in its direct vicinity are one of the crucial factors
with regard to operational safety of the dump. It can
be said that the hydrogeological conditions at the site
are fairly complicated due to the complex nature of
the geological environment with flows of underground
water. Surveys had done show existence of several
continuous water-saturated aquifers of underground
water. The highest main aquifer (H) is an aquifer with
a free water level, fairly sensitive to precipitation.
Other two aquifers are more permeable, hydraulically
fairly isolated sandy spots in deeper zones of the Sarmatian basement. The underground water has lifting
characteristics and its piezometric level reaches 1.0
to 2.0 m above the free level of the H aquifer. The
original hydrogeological conditions were significantly
CONSTRUCTIONAL AND TECHNICAL

CHARACTERISTICS OF THE REPOSITORY
At present, the Republic Radioactive Waste Dump

consists of two double-lines of storage boxes made
from reinforced concrete. Four boxes are grouped
into one dilatation block with layout dimensions of
18.6 6.0 m and height of 5.5 m, placed on a common
foundation slab 0.6 m thick. There are five dilatation
blocks in one line. An expansion gap between the
blocks is planned to be 50.0 mm thick. The storage
boxes are covered by 0.5 m thick panels from reinforced concrete. A system of two drainages (control
drainage and monitored drainage) is built at the bottom of the storage boxes and under the foundation.
The main goal of the dump drainage system is to
ensure inspection and safe removal of potential leaks
of underground and surface water contaminated by
radioactive substances from the inner space of the storage boxes into the basement and/or from the basement
into the storage boxes. Side walls and the bottom of the
monolithic storage boxes from reinforced concrete are
protected by a layer of clayish soil. A 1.0 m thick layer
of compacted clayish soil is located under the tanks
to ensure waterproof basement for the storage boxes.
Placed on this layer is a 0.6 m thick layer of sandy
gravel. Side walls are protected by a 3.5 m thick vertical clay sealing linked with a horizontal sealing soil
layer, forming a compact unit the so-called clay sealing tub. The double-lines are protected against climatic
effects by a mobile steel hall when the storage boxes
are filled with waste. Considering the general reliability of RRWD, the structures of the storage boxes are
the most important and decisive building structures.
At the same time, these structures will face the highest demands due to effects of forces, deformation and
radiation. A typical cross and longitudinal section of
the 1st and 2nd double-line is shown on Figure 3 and
Figure 4.
326
Figure 4. Typical longitudinal section of the 1st double-line

of storage boxes in RRWD in Mochovce.
4 ASSESSING STABILITY IN RRWD

ACCORDING TO LIMIT STATES
In order to make a general assessment of RRWD
safety and usability, it is necessary to examine a set
of geotechnical issues predetermined by the challenging nature of building structures at the dump and by
complex engineering-geological conditions. In terms
of compliance with requirements according to group
I and II of limit states, solutions for the problems can
be divided into two main sets as follows:
Stability calculations to assess overall stability of
RRWD and the adjacent rock environment affected
by the construction in specific stages of construction, operation and expected lifetime of the repository, including basement for the storage boxes.
Deformation calculations to assess final and
uneven settlement of decisive building structures
of the dump, taking into account specific stages of
construction, operation and expected lifetime of the
dump.
To solve the first set of problems related to security
and operative capacity of RRWD in Mochovce in every
construction stage and after the construction, it was
necessary to make a series of stability calculations to
determine:
Stability of the 1st and 2nd double-line,
Stability of adjacent slopes,
Bearing capacity of reinforced concrete boxes.
The second set of problems deals with how to ensure
usability and smooth operation of the building structures and technological units at the dump in terms of
size of the final and uneven settlement of the storage boxes. A requirement to assess horizontal shifts of
storage box walls at places of expansion gaps, planned
to be 50.0 mm wide, stems out from the structuraltechnical solution of decisive bearing structures,
including bearing walls of the storage boxes. Considering the height of storage box walls, even a small uneven
settlement or inclination of tough reinforced concrete
structures of the storage boxes can cause their damage,
resulting in a loss of their planned function. The extent
of settlement and uneven settlement of the repository
basement must be known to assess also function of
the drainage and sealing systems located in revision
and inspection shafts. In case of drainage systems, this
involves especially the required gradient needed for a

gravitational drainage of leaked water. The structural
design of the dump shows a concentrated intensity of
load in the geometric centre of the double-lines with
storage boxes, so the method and system of placing
fiber-concrete containers into the storage boxes of the
dump is very important.
The effects of uneven settlement of the basement
result in formation of local depressions in the drainage
system, so their thickness must follow the extent of
the uneven settlement of the basement to ensure the
required gradient for gravitational water drainage in
the long term. Occurrence of local depressions would
result in higher and permanent hydraulic load of specific sealing layers in the dump, affecting physical
properties of the soil in the sealing layer, in particular
the state of consistency and then strength and deformation properties. Such changes would result in higher
permeability of the sealing tub, higher compressibility and lower bearing capacity of the RRWD subsoil.
Localization of maximum tensions and deformations
given by effects of external and deforming load can
indicate potential defect areas that must be taken into
account in the design of structures of specific layers
for the final coverage.
The most demanding and the most important task
in the process of addressing these problems is to
transform actual material properties of the natural
rock to an idealized calculation model that must realistically depict the actual behavior of the assessed
structures. It is therefore necessary to pay proper attention to this task. This issue is complicated even more
by significant non-homogeneity of the natural rock
environment, but primarily by the fact that material
properties of fine-grain soil in the subsoil are very
variable in time and tension. It is very difficult to
define complex constitutional relationships in calculation models that would truly describe the actual
processes in the soil massif resulting from effects of
the load. This task can be simplified to a certain degree
by a reasonable idealization (simplification) of actual
soil properties, respecting the inevitable requirement
to preserve the physical essence. It is therefore necessary to determine physical, strength and deforming
properties of the basement soil and of the soil incorporated in the sealing tub based on standard laboratory
test on integral and undamaged samples in accordance
with applicable standards.
To define calculation models and limit conditions
of specific tasks, it is necessary to define the following
input data:
327
Spatial arrangement and geometrical shape of

building structures and construction at the area
concerned;
Extent and distribution of permanent, incidental,
extraordinary loads and their combinations;
Figure 5. Calculation model divided into finite elements.

Figure 6. Isoareas of subsoil settlement for storage boxes
calculated for the Load State 2.
Physical, strength and deforming properties of artificial building materials (concrete and reinforcedconcrete structures);
Physical, strength and deforming properties of the
natural rock environment at the area of concern and
of the soil incorporated in RRWD.
In order to assess general reliability of the dump, it is
not permissible to exceed any limit state and to allow
deformation of any structural unit during the whole
period of active operation of the repository (organization of filling the storage boxes by fiber-concrete
containers) and during the period of the planned and/or
actual lifetime of the dump once it is finally covered. Specific calculations must therefore depict all
the crucial load states that occur or might occur in
the future. The extent of load on specific building
structures and subsoil in RRWD and effects of the
load on the environment is a very variable quantity
in terms of time. It will depend mostly on the construction process applied for the main dump building
structures, schedule of organized filling of the storage boxes by fiber-concrete containers during active
operation of the dump and constructional stage of
the final dump coverage. The intensiveness of load
will therefore depend on a progressive increase in
permanent, long-term and short-term, incidental and
extraordinary load and their combinations.
Mathematical modeling using the method of finite
elements was used to calculate tensions and deformations in the RRWD subsoil. The spatial model for
the model calculations was developed according to
assumptions of the linearly elastic half-space theory
(Figure 5).
Basic parameters of the geotechnical model arise
from an engineering-geological survey and structural
design of the dump. Geotechnical calculations define
four load states (LS) describing the following crucial

stages of dump construction and operation:
LS 1 Load on the subsoil due to own weight of
the reinforced-concrete storage boxes,
LS 2 Load after filling the 1st double line with
waste containers,
LS 3 Load after filling the 2nd double line with
waste containers,
LS 4 Load after closing the 1st and 2nd double
lines and the final coverage of the dump.
Resultant settlement (vertical displacement) of the
subsoil under the storage boxes calculated for the Load
State 2 is shown on Figure 6.
The factor of time is very important to address these
problems, because the expected lifetime for RRWD
was determined by the Slovak Nuclear Supervision
Office to be 300 years based on the period of the
institutional inspection of the dump. Consolidation
processes in the dump basement and progressive redistribution of forces occurring in dependence on the rate
of loading the foundation gap (filling the storage boxes
by containers) are closely related to the time factor.
Calculations on forecasts make it possible to design a
suitable method of container placement, with favorable
effects on the overall average settlement and uneven
deformations that will have been developed depending on time in couples of decades. Stability of the
Republic Radioactive Waste dump must be ensured
with required reliability during the whole term of its
lifetime.
5
CONCLUSION
Construction of specific buildings intended to store

low and medium radioactive waste is an extremely
328
challenging task for engineers. Given the demands

and required reliability of storing radioactive waste
in RRWD Mochovce, the issue must be discussed in
its complexity. The time aspect is a significant factor
affecting a complex approach to this problem and specific periods of active operation and planned lifetime
of the dump must be taken into account. To formulate
specific geotechnical problems, it is necessary to focus
primarily on assessment of structural works in terms
of safety and usability of building structures and technological units. Much emphasis should be given on
definitions of limit conditions for specific tasks with
focus on their variability in time. A correct definition
of interactions between the subsoil and the building
structure to ensure the required level of environmental
protection is an essential requirement to address such

a complex interdisciplinary issue.
ACKNOWLEDGEMENT
This paper was developed in terms of a grant task of the
Slovak Republic Ministry of Education No 1/2135/05.
REFERENCES
Kuzma, J. et al. 1998. ExperimentalAssessment of the Republic Radioactive Waste Dump in Mochovce. Bratislava.
329
Critical ventilation velocity in large cross-section road tunnel fire

Z.X. Li, X. Han & K.S. Wang
Shanghai Institute of Disaster Prevention and Relief, Tongji University, Shanghai, P.R. China
ABSTRACT: Based on CFD method, this paper conducted simulation analysis of lorry and car fires for large
cross-section road tunnel. The smoke distribution feature was discussed and the corresponding results of critical
ventilation velocity were also compared with the values calculated by related empirical formula. It would be
helpful to work out a ventilation system strategy for the road tunnel.
INTRODUCTION
Recent incidents however have drawn widespread

attention to the risks of fires in road, rail and masstransit tunnels (e.g. Kings Cross, Channel Tunnel,
Mont Blanc, Tauern, Gotthard). The incidents have
resulted in casualties as well as large direct and indirect
economic damage. The control of smoke flow during
a road tunnel fire is often an important part of fire
safety measure. On the other hand, CFD (Computational Fluid Dynamics) simulations are increasingly
used to estimate the effects of fires in road tunnels.
However, little work has been done to analyze comprehensive fire characteristics of the large cross-section
road tunnel. On the basis of CFD method and with
the help of FDS simulation software, this paper conducted simulation analysis of lorry and car fires for
large cross-section road tunnel. Two fire scenarios
were constructed and five different kinds of longitudinal ventilation conditions were considered. The
smoke distribution feature was discussed and the corresponding results of critical ventilation velocity were
also compared with the values calculated by related
empirical formula. It would be helpful to work out a
ventilation system strategy for the road tunnel.
2
2.1
CFD SIMULATION SCENARIOS

Brief introduction of FDS
In these simulations, FDS (Fire Dynamics Simulator)

4.06 which was released by NIST (National Institute of Standards and Technology, USA) was used.
FDS is a Computational Fluid Dynamics (CFD) model
with LES (Large Eddy Simulation) of fire-driven fluid
flow. The model solves numerically a form of the
Navier-Stokes equations appropriate for low-speed,
Figure 1. Cross-section of road tunnel.
thermally-driven flow with an emphasis on smoke and

heat transport from fires. The partial derivatives of
the conservation equations of mass, momentum and
energy are approximated as finite differences, and the
solution is updated in time on a three-dimensional,
rectilinear grid. Thermal radiation is computed using
a finite volume technique on the same grid as the flow
solver.
2.2 The design of the fire scenario
As for large cross-section road tunnel, the fire smoke
distribution feature under different ventilation situation was simulated. One road tunnel which has
100 m length and 15 m diameter was constructed as
331
Figure 4. Heat release rate of fire source in scenario 1.

Figure 2. CFD model in scenario 1.
Figure 3. CFD model in scenario 2.
Figure 5. Heat release rate of fire source in scenario 2.
simulation model, as shown in Figure 1, Figure 2 and

Figure 3. One lorry which designed as fire source was
located at 37 m from the upstream ventilation crosssection and 63 m from the downstream ventilation
cross-section, as shown in Figure 2. The lorry was supposed to be on fire first, and the heat amount released
by the lorry would emblaze a car near the lorry. Two
fire scenarios were set up, including: (1) scenario 1:
the car at the right of the lorry; (2) scenario 2: the car in
front of the lorry. Both scenarios involved five different kinds of longitudinal ventilation conditions. The
ventilation velocities were set as 2 m/s, 2.5 m/s, 3 m/s,
3.5 m/s, and 4 m/s respectively.
During the simulation, the lorry was burning first.
Then at a certain time, when the heat amount the car
received exceeding 16 kW/m2 , the car would begin to
burn. The heat release rate of the lorry and the car were
20 MW and 5 MW respectively. The total heat release
rate was shown in Figure 4 and Figure 5 separately.
3 ANALYSIS OF CFD SIMULATION RESULTS

Figure 6. The smoke distribution feature under ventilation
velocities of 2 m/s, 2.5 m/s, 3 m/s and 3.5 m/s respectively in
scenario 2.
3.1 The brief analysis of the simulation results

Since the heat release rates of two scenarios were
almost of same value, the simulation results demonstrated that the relative position of the lorry and the
car didnt distinctively affect the smoke distribution
feature. As for the same environmental conditions,
there was not much difference between these two
fire scenarios, as shown in Figure 6, Figure 7 and

Figure 8 respectively. For example, under the same
longitudinal ventilation condition, the smoke density
at the location of 10 m upstream the fire source was of
similar distribution type.
332
Figure 7. Smoke density at the location of 10 m upstream

the fire source in scenario 1.
Figure 8. Smoke density at the location of 10 m upstream

the fire source in scenario 2.
3.2 Obtain of the critical ventilation velocity

The simulation results illustrated that the road tunnel fire yielded high production of smoke. As a result
of fire temperature, the smoke would have a lower relative density than the surrounding air in the tunnel
and therefore rose to the ceiling of the tunnel. Consequently the smoke would remove itself away from the
fire source in a layer along the ceiling. Along the bottom the cold air would be sucked towards the fire. This
created two flowing layers, i.e., the hot smoke layer and
the cold smoke-free layer of air. The phenomenon was
called stratification.
The hot smoke layer gradually formed as the hot
smoke curled upwards. With continuous yielding of
smoke from the fire source, the hot smoke layer
swelled rapidly and spread towards both sides of the
tunnel. Meanwhile, the cold air from the lower part of
the tunnel flew to the fire source. In this way, the symmetry re-circulated air was formed on both sides of
the fire field. When the longitudinal ventilation was
started, the smoke on both sides of the fire source
emerged asymmetry. If the ventilation velocity is quite
small, it couldnt prevent the smoke from diffusing in
the direction opposite to the ventilation stream. This
situation, namely smoke flows against the ventilation,
is adverse to the prevention of the smoke spread (the
hot smoke maybe emblaze the traffic in the upper side
of the fire) and would be harmful to the safety of the
firemen. In order to prevent the backflow of the smoke,
the longitudinal ventilation velocity should be greater
than the critical one.
As ventilation velocities were set as 2 m/s and
3 m/s, and at the location of 10 m upstream the
fire source, the maximum smoke density would
approached 250 mg/m3 and 280 mg/m3 separately.
While the ventilation velocity was set as 4 m/s, the
corresponding smoke density almost kept zero. It obviously indicated that the critical ventilation velocity
went between 3 m/s and 4 m/s.
Oka and Atkinson had proposed a method to calculate the critical velocity, which was to get the smoke
backflow distance in different ventilation through the
detectors first. Then the value of the distance was
placed in the reference frame for backflow distance
and ventilation velocity. To joint these points in the
reference frame and extend the line to the y-axis,
the y-axis value of the point of intersection was just
the critical velocity. This paper would also obtain the
critical ventilation velocity in this way.
The spread of the smoke under the ventilation velocities of 2 m/s, 2.5 m/s, 3 m/s and 3.5 m/s were presented
in above mentioned figures. These figures could give
out the related distance values which the smoke spread
upstream the fire source. Based on the detectors, we
could get the backflow distance in four situations of
120 m, 80 m, 48 m and 10 m. These numerical values were put up in the reference frame for backflow
distance-ventilation velocity. Jointing the points and
extending the line in accordance with the method
referred previously, then the critical ventilation velocity could be obtained as 3.65 m/s.
3.3 Comparative analysis of the critical ventilation
velocity
Oka and Atkinson had set up a formula for calculation
of tunnel critical ventilation velocity as follows:
333
parameter were of the same meaning as formula

(1). Through formula (5), we could obtain the critical ventilation velocity as 2.893 m/s. Comparing this
value with the simulation one, the relative error was
20.7%.
Figure 9. Critical ventilation velocity for large cross-section

road tunnel.
where Q = the heat release rate, kw; Vcritical = the critical velocity, m/s; Q = dimensionless heat release
rate, and V = the dimensionless critical velocity,
0 = ambient air density, kg/m3 ; CP = specific heat
capacity, kJ kg1 K1 ; T0 = ambient temperature, K;
g = acceleration due to gravity, m/s2 ; H = height from
the surface of fire source to tunnel ceiling, m. Using
these values we could obtain the critical ventilation
velocity of the tunnel as 3.208 m/s. Comparing the
simulation value with the one given out by formula
(1), the relative error was 12.12%.
Wu and Bakar further amended the formula
(1). They considered the influence of the breadth
of the tunnel and replaced the height of the tunnel
with the hydraulic tunnel height H , which was defined
as the ratio of 4 times the cross-sectional area to the
tunnel wetted perimeter. Thus, the new formula was
illustrated as follows:
where Q = dimensionless heat release; V =

dimensionless critical ventilation velocity, other
CONCLUSION
The simulation results demonstrated that specifying a

critical ventilation velocity was helpful to prevent back
layering of smoke during a tunnel fire. Hence, it was
very important for smoke management of large crosssection road tunnel. In order to find cost-effective
methods to upgrade fire safety in existing tunnels or
tunnels to be built, the practical tests as well as simulation analysis should be further carried out. Hence, the
corresponding performance-based calculation method
could be put up to improve the computation of critical ventilation velocity. Meanwhile, more influencing
factors need to be considered, such as the impact
of road tunnel physical characteristics and the traffic
capacity, etc.
ACKNOWLEDGMENTS
The support of the Natural Science Foundation of
China (Grant No. 50678124) is gratefully appreciated.
REFERENCES
Bettis, RJ., Jagger, SF., Wu, Y. Interim validation of tunnel
fire consequence models; summary of phase 2 tests.The
Health and Safety Laboratory Report IR/L/FR/93/11, The
Health and Safety Executive. UK, 1993.
Danziger, NH., Kennedy, WD. Longitudinal ventilation analysis for the Glenwood canyon tunnels. Proceedings of
the Fourth International Symposium Aerodynamics and
Ventilation of Vehicle Tunnels. York, UK, 1982: 169186.
Heselden, AJM. Studies of fire and smoke behavior relevant to tunnels. Proceedings of the Second International
Symposium of Aerodynamics and Ventilation of Vehicle
Tunnels. Paper J1, 1976.
Kennedy, WD., Parsons, B. Critical velocity: past, present and
future. Paper presented in the One Day Seminar of Smoke
and Critical Velocity in Tunnels. London, 2 April 1996.
Oka, Y., Atkinson, GT. Control of smoke flow in tunnel fires.
Fire Safety Journal. 1995; 25(4): 305322.
Wu, Y., Bakar, MZA. Control of smoke flow in tunnel fires
using longitudinal ventilation systems-a study of critical.
Fire Safety Journal. 2000; 35(4): 363390.
334
Metro tunnels in Buenos Aires: Design and construction procedures

19982007
A.O. Sfriso
Department of Estabilidad, University of Buenos Aires, Argentina
ABSTRACT: The City of Buenos Aires, Argentina, is expanding its metro network. Some 13 km of new tunnels
have been excavated since 1998 and some 20 km are scheduled for construction in the near future. Many major
improvements have been implemmented during these years in the fields of design and construction procedures.
Some of the achievements and lessons learned are described in this paper, including: characterization of Buenos
Aires soils for the numerical modeling of NATM tunneling, description of the design and construction procedures
in use and some comments on the observed ground behavior during construction.
INTRODUCTION
The City of Buenos Aires is extending its metro

network as shown in Figure 1. On-going projects
are: Line A, extended 4 km, Line B, extended 4 km;
Line E, extended 2 km; and new Line H, 5 km long.
Some 20 km of new Lines F, G, I are scheduled for
construction in the near future (SBASE 2006).
Landmarks of new construction procedures are: i)
introduction of shotcrete, Line B, 1998 (Fig. 2); ii)
belgian tunneling method, Line H, 2000 (Fig. 3); iii)

full face excavation, Line B, 2004 (Fig. 4).
Geotechnical and structural analysis techniques
evolved concurrently, from earth-load theory to state
Figure 2. First use of shotcrete, Line B.
Figure 1. Metro network in Buenos Aires. Existing (A, B,

C, D, E, H) and new projects (F, G, I).
Figure 3. Belgian tunneling method, Line H, 2000.
335
Table 1. Assumed in-depth variation of K0 .
0
8/12
20/24
Table 2.
0
8/12
12/14
20/24
Figure 4. Full face excavation, Line B, 2004.
of the art computer simulation of construction procedures and calibration of constitutive models via back
analysis of monitoring data (Nez 1996, Sfriso 1996,
1999, 2006).
2
CHARACTERIZATION OF BUENOS AIRES

SOILS FOR TUNNELING
Buenos Aires City soils have been described in other

contributions (Bolognesi 1975, Fidalgo 1975, Nez
1986a, 1986b). Briefly, the Pampeano formation
underlying BuenosAires is a modified Loess, overconsolidated by dessication and cemented with calcium
carbonate in nodule and matrix impregnation forms.
Except for the heaved upper three to six meters, penetration resistance is systematically NSPT > 20 with
some heavily cemented zones that exhibit very weak
rock behavior with NSPT > 50 (Nuez 1986b).
The most used site investigation technique in
Buenos Aires is SPT penetration using a 2 sampler along with standard lab testing and CTUC testing on recovered samples. Some plate load testing
and Menard pressuremeter testing have been recently
included as part of the field investigation specifications for metro projects (Sfriso 2006).
2.1
Underground construction in the Pampeano

Formation
The Pampeano formation is very favourable for underground construction due to its high stiffness, reliable compressive strength, rapid drainage and good
frictional behavior when drained.
Two particular characteristics of the formation must
be accounted for in the design of underground projects:
i) the Pampeano formation is fissured and has lenses
of quasi-granular behavior, forcing the installation of
a primary support close to the face in order to avoid
Depth m
K0
to 8/12
to 20/24
to 30/32
0.550.70
0.651.00
0.550.80
PLT Modulus of subgrade reaction.

Depth
m
Primary loading
MN/m3
Un-Reloading
MN/m3
to 8/12
to 12/14
to 20/24
to 30/32
200300
400600
600800
250500
500800
8001200
12001800
6001400
crown overexcavation; and ii) materials drain at a speed

comparable to that of the construction.
Due to these factors, the max allowable drift without
support is about 2.5 meters. Up to this maximum, the
unsupported drift has very little influence on the resulting settlements, as soil behavior remains quasi-elastic
(Sfriso 2006, Nez 2007).
2.2 In situ stresses
It is accepted (Bolognesi 1991, Nez 1986a, 1986b,
Sfriso 1999, 2006) that upper Pampeano soils are
overconsolidated by dessication to an equivalent pressure 0.81.2 MPa. Table 1 lists the assumed in-depth
variation of K0 used for the design of underground
structures (Sfriso 2006). These figures have not been
actually measured directly but estimated after backanalysis of monitoring data.
2.3 Modulus of subgrade reaction
The most reliable information of in-situ stiffness is
retrieved via plate loading tests performed in vertical
shafts or pilot tunnels. Typical in-depth variation of the
modulus of subgrade reaction, as determined by PLT,
is listed in Table 2.
2.4 Parameters for numerical modelling
Hyperbolic model (Duncan 1970, Vermeer 1998)
has been extensively used for the numerical analysis of underground construction in Buenos Aires soils
(Sfriso 1999, 2006). After eight years of continuous
usage and calibration, a set of input parameters for
the Plaxis implemmentation of the hyperbolic model
(Vermeer 1998) has been found to best represent
336
Table 3.
Material parameters used for numerical simulations.
cu (KPa)
u ( )
c (KPa)
( )
( )
Er50 (MPa)
Erur (MPa)
m ()
()
Rf ()
Fill
min
max
08/12
min
max
8/1220/24
min
max
> 20/24
min
max
20
8
0
28
0
10
25
0
0.20
0.85
50
15
5
30
0
20
50
0
0.20
0.90
50
10
10
28
0
60
150
0
0.20
0.80
100
20
25
31
3
100
250
0
0.30
0.90
110
5
25
30
0
75
180
0
0.20
0.80
220
20
50
34
6
150
300
0
0.30
0.90
40
0
15
28
0
60
140
0
0.25
0.80
120
5
30
31
3
100
220
0
0.35
0.90
the observed behavior of tunnels, caverns and open

pit excavations. This set is listed in Table 3. Stressstrain relationship of the HSM model is reproduced in
Equations 1a to 1d.
Figure 5. Cross section of a typical two lane, full face

tunnel, lines B and H.
In Equations 1a to 1d and Table 3, 1 and 3 are the

major and minor principal stresses, 1 is the major
principal strain, c is either undrained cohesion cu or
drained cohesion c , is either undrained friction angle
u or drained friction angle , is dilatancy angle,
r
r
E50
and Eur
are reference loading/unloading Youngs
modulus, m is stiffness exponent, is Poissons ratio
and Rf is the failure ratio.
CONSTRUCTION PROCEDURES
3.1 Tunnels
Construction procedures evolved from german
method (Fig. 2) to belgian method (Fig. 3) and have
probably reached an optimal stage with full face excavation (Fig. 4). Figure 5 shows the cross section of a
typical two lane, full face tunnel, as used in Lines B
and H. A 15 cm unreinforced shotcrete layer and 1.0 m
spaced lightweight lattice girders account for the primary support of the tunnel, later supplemented with
3040 cm of cast-in-place unreinforced concrete.
The Metro authority requires that tunnels remain

dry during operation, thus rendering cast-in place secondary lining as the cheapest option, when compared
to membrane barriers and secondary shotcrete lining.
Some full face, all-shotcrete sections with impermeabilization barriers have been successfully built in
lines H and B.
No closure of the structural ring is needed for stability, and therefore advance rates of 2.5 m3.5 m per
12 hr shift are consistently achieved. After the tunnel
is excavated, a cast in place invert is placed in 5 m6 m
segments, allowing for the placement of the secondary
lining in single poured 5 m segments. Figure 6 shows
a tunnel after placement of the invert, while Figure 7
shows the formwork being driven into the tunnel.
3.2
Underground stations
Underground caverns for metro stations have been

built using many techniques including: i) cut & cover
slab-on-piles; ii) underground excavated main cavern
& open pit excavated upper hall; and, iii) underground
excavated main cavern & upper hall.
337
Figure 6. Tunnel after placement of the invert, Line B.
Figure 8. Echeverra Station, Line B. German tunnelling

method.
Figure 7. Formwork used to cast the secondary lining,

Line B.
Figure 9. Corrientes Station, Line H. Full face excavation.
As per 2007, three stations are in excavation stage.

The first two are Echeverra Station (Fig. 8) and Villa
Urquiza Station, Line B, where the german tunnelling
method was used. The third station is Corrientes Station, Line H, where full face excavation is being used
(Fig. 9).
Corrientes Station is the latest and more challenging improvement to construction procedures used in
Buenos Aires metro tunnelling so far. It is an underground cavern 14.1 m high, 18.9 m wide and 135 m
long (Fig. 10). On top of the main cavern, a 6 m
high access hall shall be excavated full-face after
completion of the secondary lining of the main cavern.
The primary lining of Corrientes Station is formed
by 2040 cm mesh reinforced shotcrete placed in two
layers, and 1.0 m spaced lightweight lattice girders.
The construction procedure is shown in Figure 11.
Full-face excavation is accomplished via a series of
four benches, each one 5 m long. Two excavators are
permanently set at the two top and two bottom benches,
respectively. The bottom bench excavator alternatively
lies on soil or on top of the cast-in-place invert,
included into the primary support lining to reduce
Figure 10. Longitudinal sketch of Corrientes Station,

Line H.
costs and time schedule. While excavation of Echeverra Station took some 14 months, it shall took some 6
months to complete Corrientes Station, with advance
rates 1.0 m per day.
338
Structural forces obtained with equations 2a, 2b and 2c

compare within 10%15% with those computed using
the more involved procedure by Einstein & Schwarz
(Einstein 1979).
4.2 Simulation of construction procedures
Construction procedures are simulated using 3D
elastoplastic models that allow for the estimation of
surface settlements, the computation of face stability
and the determination of structural forces acting on
the primary lining. Structural forces computed with
3D FEM are some 20% lower than those obtained with
Eqns. 2a, 2b, 2c. These equations, when applied to the
tunnel shown in Fig. 5, resulted in M = 0.55 KNm/m
and N = 416 KN/m. 3D numerical models yielded
M = 0.47 KNm/m and N = 415 KN/m. (Sfriso 2006,
Nez 2007).
4.3
Secondary lining
Metro authority requires that the secondary lining be

designed using earth-load procedures and beam on
springs analyses. Both primary lining and the effect of
construction procedures are disregarded in the design
of the secondary lining.
5
Figure 11. Construction procedure for Corrientes Station,
Line H.
4
4.1
DESIGN PROCEDURES
Primary lining
The preliminary design of the primary lining is largely

based on experience. By the time the tunnel shown in
Fig. 2 was being analyzed, a simplified design method
was developed to estimate forces acting in the crown of
the primary support of circular sections (Nez 1996).
The expressions are
where N is the normal force at crown, M is the flexure

moment at crown, pv is the vertical pressure on the
crown, D is the tunnel diameter, A is the unsupported
drift, E, are the elastic parameters of the soil mass and
Er , r are the elastic parameters of the support system.
339
GROUND BEHAVIOR
Ground behavior has been largely elastic for all construction procedures and underground structures built
so far. Disturbance to surrounding structures and facilities has always been minimal, and surface settlements
in the range 2 mm8 mm for tunnels and 4 mm15 mm
for underground caverns have been observed for all
construction procedures and soil covers. While this is
a desirable behavior from the point of view of construction and safety, it also means that uncertainty of
the predictions remain high, because it is unknown
how safe the construction procedures really are.
A numerical excercise has been performed to compare the construction procedures for safety and impact
to surroundings. A tunnel section 10 m wide, 8 m high
with a soil cover of 5 m was used, and the low side
parameters listed in Table 3 were adopted. The results
are listed in Table 4 (Sfriso 2006). It can be noticed
that the german method proved to be the least safe
construction method, due to the low safety of the
unsupported access tunnels excavated to build the side
walls (Fig. 12).
Echeverra Station and Corrientes Station have
proven enlightening experiences for the purpose of
checking predictions. For safety considerations, it was
decided that Echeverra Station be excavated using
the german method, and a max. surface settlement of
15 mm was predicted. After a series of small access
Table 4. Numerical comparison between construction procedures for tunnels.

Belgian
Full
face
4.9
4.3
4.6
7.4
5.3
6.7
0.26
0.22
0.18
0.30
0.26
0.26
German
Max. surface settlement,
undrained parameters (mm)
Max. surface settlement,
drained parameters (mm)
Max. angular distorsion,
undrained parameters (103 )
Max. angular distorsion,
drained parameters (103 )
Factor of safety,
undrained parameters ()
2.6
>7
4.7
Figure 13. Strut supported access tunnels, Echeverra Station, Line B.
responsible for the small settlements observed. Being

the first large closed ring structure ever built in Buenos
Aires, Corrientes Station is the first opportunity to
properly calibrate the unloading Youngs modulus and
the effect of soil rebound.
6
Figure 12. Unsupported pilot tunnel for side walls, german

method of tunnelling.
tunnels were excavated to improve the construction

schedule, and before the main cavern excavation was
started, a surprisingly high 10 mm settlement was
observed at surface.
When the numerical model was re-run with the
access tunnels included but without any change in
material parameters, the observed surface settlement
could be reproduced. It turned out that the unconfinement produced by the too many intersecting small
tunnels was responsible for the undesired behavior and
yielded a temporarily unsafe condition. It was dediced
that the abandoned access tunnels be supported by
struts (Fig. 13).
At Corrientes Station, the observed surface settlement 5 mm8 mm is much lower than the predicted
value of 20 mm. After interpretation of the monitoring data, it has been concluded that the unload
Youngs modulus of Pampeano soils is lower than originally estimated, and that the deposit rebound is partly
REMAINING CHALLENGES
The advancement in design and construction procedures is an endless activity. Despite the efficient
methods actually in use, some remaining challenges
need to be addressed in the near future. These are: i)
the implemmentation of a reliable procedure to measure K0 ; ii) the abandonment of cast-in-place concrete
and dry tunnels; iii) the use of robot-placed, fiber
reinforced shotcrete; iv) the implemmentation of more
advanced topographic guiding systems; v) optimizations in the usage of lattice girders; and vi) better
control of ground water during construction.
7
CONCLUSIONS
13 km of metro tunnels have been excavated in Buenos

Aires in the period 19982007. Construction procedures in 2007 include shotcrete and full face excavation both in tunnels and caverns, while design
procedures include state of the art numerical simulation of construction processes. Best fit parameters for
the constitutive models used were introduced and some
observed features of soil behavior have been described.
The Pampeano formation underlying Buenos Aires
City is very favourable for underground excavation
due to its high stiffness, reliable compressive strength,
rapid drainage and good frictional behavior when
drained.
340
Ground behavior has been largely elastic for all construction procedures and underground structures built
so far. Disturbance to surrounding structures and facilities has always been minimal, and surface settlements
in the range 2 mm8 mm for tunnels and 4 mm15 mm
for underground caverns have been observed for all
construction procedures and soil covers.
Corrientes Station is the latest improvement to construction procedures used in Buenos Aires metro tunnelling so far. An underground cavern 135 m long shall
be completely excavated in six months with surface
settlements less than 10 mm and minimal disturbance
to surroundings.
ACKNOWLEDGEMENTS
The working teams from the companies involved in
these projects, Techint, Dycasa, Roggio and SBASE,
shared their impressions, opinions and experience with
the writer, allowing him to learn many things from
them. Eduardo Nez taught the writer the basics, the
involved theories and the tricks of tunnelling, carefully
supervised the writers early works, and remains as the
day-by-day writers source for expert advice.
Bolognesi, A. and Vard, O. 1991. Subterrneos en Buenos

Aires. IX PCSMFE, Chile, III:13291350.
Duncan, J. and Chang, C. 1970. Nonlinear analysis of stress
and strain in soils. JSMFD, ASCE, 96, SM5, 16291653.
Einstein, H. and Schwartz, C. 1979. Simplified analysis for
tunnel supports. JGED, ASCE, 105, 4, 499518.
Fidalgo, F., De Francesco, F. and Pascual, R. 1975. Geologa
superficial de la llanura Bonaerense. VI Arg. Geol. Conf,
110147.
Nez, E. 1986a. Panel Report: Geotechnical conditions in
Buenos Aires City. V Conf. IAEG, 26232630.
Nez, E. and Mucucci, C. 1986b. Cemented preconsolidated
soils as very weak rocks. V Conf. IAEG, 403410.
Nez, E. 1996. Tneles de seccin circular en la formacin
pampeano. bull. SAMS, 29, 115.
Nez, E. 2007. Casagrande lecture: Uncertainties and
Approximations in Geotechnics.XIII PCSMGE,Venezuela.
In press.
SBASE, 2006. Institutional web page. www.sbase.com.ar.
Sfriso, A., 1996. Tneles de seccin circular en la formacin
Pampeano comentario. bull. SAMS, No. 29, 1619.
Sfriso, A. 1999. Tunnels in Buenos Aires: Application of
numerical methods to the structural design of linings. XI
PCSMGE, Brasil, 637642.
Sfriso, A. 2006, Algunos procedimientos constructivos para
la ejecucin de tneles urbanos. XIII CAMSIG, Argentina,
117.
Vermeer, P. 1998. Plaxis Users Manual. Balkema, Rotterdam.
Ne, 577 p.
REFERENCES
Bolognesi, A. 1975. Compresibilidad de los suelos de la Formacin Pampeano. V PCSMFE, Argentina, V: 255302.
341
Study on the earth pressure distribution of excavation chamber in EPB

tunneling
T.T. Song & S.H. Zhou
Department of Geotechnical Engineering, Tongji University, Shanghai, P.R. Chnia
ABSTRACT: In shield tunneling, the excavation stability is very important. The balance pressure which supports the work face is offered by the the mucks in the earth chamber. The ideal supporting pressure on the
excavation face is trapezoidal. In fact, the pressure is irregular. So, the conceptions of earth pressure supporting
ratio (EPSR), regular modulus of earth pressure and buffer ability of chamber are put forward. Earth pressures in
two different situations of shield tunneling in clay ground and cobble sand ground are studied. The studies show
that EPSR and regular modulus of earth pressure in soft ground are both better than those in cobble ground.
EPSR and regular modulus of earth pressure are two appropriate assessing indexes for EPB tunneling.
INTRODUCTION
With the advantages of surrounding influence, excavation speed, structure quality, working environment etc.,
the shield tunneling method is popularly applied in the
tunnels of metro, railway, road, municipal engineering
and so on.
Earth pressure balance machine (EPB) and slurry
shield tunneling are the two mostly common method in
metro tunnels. Comparing with slurry shield method,
the EPB has the merits of small construction yard, less
cost, simple technology. So the EPB is more popular
in metro tunnels.
EPB machine applies pressure to working face by
the mucks which are cut from the face sometimes with
injection of soil conditioning additives. With the pressure the working face can maintain stable. The stability
of working face is a key factor in the EPB tunneling.
The accidents brought by destabilization of working
face are the main accidents according to statistics (Qin
Jianshe, 2005).
Present researches about the shield tunneling pressure mainly focus on the theoretical pressure needed
for the working face (Anagnostou G, Kovari K., 1996;
Abdul-Hamid Soubral, 2000 & Qin Jianshe, 2005).
As for how to apply the pressure and the pressure
properties are seldom mentioned and researched. In
this research, features of working face pressure, distribution of pressure in the excavation chamber and
the pressure in the clay and cobble sand strata are
researched. Some conceptions are put forward, and
principles are summarized.
CLASSICAL EPB THEORY
The excavating system of EPB machine is composed

by shield, cutter head embedded with cutters, working chamber, pressure wall, screw conveyor, and thrust
cylinders (see Fig 1). When excavation, thrust cylinders apply force to pressure wall, and the pressure wall
applies the pressure to the working face by the mucks
in the earth chamber. The pressure is used to balance
the water and soil pressure in ground.
EPB machine adjusts pressure to balance water and
soil pressure in working face by the manners of changing the pressure of thrust cylinders and the rotation
speed of screw conveyor (see Figure 1). When it is
necessary to increase the pressure in earth chamber,
the rotation speed should be decreased and the thrust
force should be increase vice versa.
Classical earth pressure assumes these:
1. The pressure applied by shield machine to the working face is trapezoidal. And the trapezoidal pressure
balances with the water and soil pressure in working
face in order to maintain facing balance.
2. The pressure applied to the ground is equal to the
pressure in the pressure wall (M. Herrenknecht &
U. Rehm, Aug, 2003). So, the working face pressure
is acquired through the pressure values shown in the
operating panel.
Obviously, the conditions for classical earth pressure are satisfied with the followings:
1. The mucks in earth chamber should have the
properties of ideal plastic fluidity.
343
Figure 1. Soil pressure + water pressure = earth pressure in working chamber (Wassmer, Treceno & ANdreossi, 2001).
2. The opening rate of cutter head is enough.

3. There is no pressure loss in the screw conveyor.
In fact, EPB cant reach this ideal state in the
practice of excavation.
3
DEFINITIONS RELATED EARTH PRESSURE
The pressure applied by shield machine is related to

the stability of working face directly. In practice, its
hard to observe and gauge the pressure of the working
face. So, some indirect ways to estimate the pressure
of working face, as below:
1 The pressure sensor in the pressure wall. This way
is thought as the most efficient way. But it has
difference between the two pressures.
2 The force of thrust cylinders and the torque of cutter head. They increases with the earth pressure
increase;
3 The volume of mucks discharge. More discharge,
less earth pressure.
For further research, the following definitions are
given:
3.1
Earth pressure supporting ratio (EPSR)
The classical earth pressure theory assumes that the

supporting pressure on the working face is supplied
by earth in the earth chamber (see Fig 1). In fact, the
supporting pressure is composed by two parts: earth in
working chamber and the plane of cutter head. As to
the Figure 2, the earth pressure is mainly transferred
though the open of cutter head. But the plane also
contributes supporting pressure. So the earth pressure
supporting ratio (EPSR) is defined as the ratio of earth
pressure in the total supporting pressure.
Figure 2. Schematic plan of open and plane.
where EPSR = Earth Pressure Supporting Ratio;

earth supporting pressure: the total pressure of earth
pressure on the working face;
total supporting pressure: the total supporting pressure on working face including earth pressure and
cutterhead plane supporting pressure;
The total supporting pressure is assumed the balance pressure on the working face.
The ratio is related to the shape of cutter head, opening rate and the control of earth pressure. The value of
the ratio is between 0 and 1 which indicates the state of
earth control when excavating. For example, the ratio
of cutter head of spoke shape is bigger than that of
plane shape.
3.2 Regular modulus of earth pressure (RMEP)
Five pressure sensors are located on the pressure wall
(see Fig 1). The position of the sensors are shown in
Figure 3, and the height is shown in Table 1.
344
Figure 4. Total pressure makeup: additive pressure of

facing + lateral pressure of mucks.
Figure 3. The Distribution of earth pressure sensors.

Table 1.
Position of the sensors.
Sensor No.
Position(m)
1
1.2
2
1.8
3
2.7
4
3.4
Here, the buffer ability of chamber is defined as the

proportion of discharged mucks in the whole chamber at the condition of screw conveyor working for 5
minutes with maximum rotation speed.
5
5.4
The regular modulus of earth pressure RMEP is

used for indicating the regularity of the earth pressure
in working chamber. The value of RMEP is defined as
related coefficients: R2 between values of earth pressure and the corresponding vertical position. The value
of the modulus is between 0 and 1 which indicates the
fitting extent of the fitted regression line. The ideal
value is 1. The value is closer to 1, better regular of the
earth pressure is.
It is deemed that the buffer ability is good if the

proportion of discharge mucks is less than 30%.
PROPERTIES OF EARTH PRESSURE

OF EPB EXCAVATION
4.1 Analysis of pressure makeup

where
Yi : actual data points of earth pressure

Y i : predicted value of the regression model
The detail definition of R2 can referred Richard A.
Johnson (1992).
3.3
Buffer ability of working chamber
It cant keep the balance pressure when the mucks

in the working chamber are over discharged due to
improper operation. The volume of the working chamber should have the ability of maintaining the pressure
due to improper discharging operation. So the chamber should buffer the earth pressure when the mucks
are sudden loss.
The pressure wall is in the behind of working chamber.

The wall undertakes the main face pressure. Here the
series S type EPB shield machine of Herrenknecht are
take an example to research the wall pressure. The
earth pressure sensors in the wall are distributed as the
Figure 3. There are five sensors in the wall. The serial
number and the position are shown in Figure 3.
The measured pressure shape is trapezoidal in most
common situation (see Fig 1 & Fig 4). The trapezoidal
pressure is composed of two parts: one is rectangular
and another is triangular. The triangular can be deemed
as the lateral pressure of the mucks, and the rectangular
can be deemed as the additive pressure applied by the
water and soil pressure of the working face.
Analysis hereinbefore, there wont be additive pressure if the pressure shape is triangular. At this situation,
if the thrust force is big and the face is stable, the cutter
head will undertake large portion of face pressure. So,
the earth pressure support ratio EPSR is low which
lead to the large contact stress between cutter head
and face and large torque of cutter wheel. Usually, in
this situation, the wear of cutters and cutter head are
serious.
345
Figure 5. Earth pressure distribution diagram (Herrenknecht & Rehm, 2003).

Table 2. The ground parameters of Guangzhou.
3
Earth pressure of the sensors
(g/cm )
w(%)
c(kPa)
( )
Es1-2(MPa)
1.95
31
0.8
22.8
18.1
4.7
R2 = 0.9761
5.0
Pressure properties in clay
4.0
Usually, the mucks of clay strata have a good plastic

fluidity which can apply a regular pressure to the excavation face. So, the earth pressure supporting ratio is
high; and the modulus of the earth pressure is high too.
The stability of working face can be well controlled if
the working chamber satisfies the mucks buffer.
Nevertheless, there is pressure loss in the screw conveyor in this kind of strata. The pressure gradient curve
is shown in Fig 4. It is inevitable that pressure falls in
the screw conveyor because the pressure is 0 in the discharging outlet. The pressure gradient depends on the
fluidity and impermeability of the mucks. The pressure in the inlet of screw conveyor has fallen to 80%
(see Fig 5). It indicates that the pressure isnt regular
due the existence of screw conveyor though the mucks
is of plastic fluidity. (see left in Fig 5)
The excavation data of ring No. 219 in one section of
line 5 of Guangzhou metro are taken for an example.
The cutter head of the EPB machine is plane type,
and opening rate is about 28%. There are two types
of cutters: disc cutter and scrape cutter. The length of
working chamber is 1 m. The position of earth pressure
sensors are shown in Figure 3. The tunnel is in the
stratum of <4 2> whose main parameters are shown
in Table 2.
The Figure 6 is the figure of earth pressure data
which are acquired from the earth pressure sensors
in pressure wall. The number order of the pressure
sensors is from the top down. They are 1#, 2#, 3#, 4#
and 5#. From the data it can obtain that the regular
modulus of earth pressure RMEP is about 0.98. This
value shows the earth pressure of pressure wall is more
regular.
4.2
6.0
3.0
2.0
1.0
0.0
0.06 0.07 0.08 0.09
0.1
0.11 0.12 0.13
MPa
Figure 6. Mean earth pressure in the pressure wall.
Table 3. Assessment of earth pressure supporting ratio.
Additive stress/ Max. stress
Actual torque/
Max. torque
<0.8
>0.8
<1.1
>1.1
0 30%
30 50%
>50%
<0.5
<0.3
0.5 0.8
0.3 0.5
>0.8
>0.5
<0.1
0.1 0.3
0.3 05
The proportion of earth pressure and cutter head

is hard to be divided. The earth pressure supporting
ratio cant be calculated precisely. So, it can be judged
by the earth pressure in the pressure wall and the
torque of cutter head by the following way.The additive
346
Table 4.
Main index of the EPB (S365).
Machiner
type
Excavation
diameter(m)
Rating
torque(KNm)
Max.
thrust(kN)
Max.
speed(mm/min)
Opening
ratio
Cutter
head
EPB TBM
6.28
5980
34210
80
28%
disk and scrape
Table 5. The indexes of the stratum.
Earth pressure of the sensors
(g/cm3 )
w(%)
c(kPa)
( )
E0 (MPa)
K(m/d)
2.2
31%
01
38
45
27
6.0
R2 =0.93
5.0
pressure can be known from the pressure wall. Because

the additive pressure is applied by face, higher additive pressure, higher earth pressure supporting ratio.
If the cutter head plate undertakes more facing pressure the torque of the cutter head will become bigger.
According to empirical summary and modification,
the method using additive pressure and torque to estimate earth pressure supporting ratio is put forward.
(shown in Table 3)
The actual torque of this ring is 3650 kNm while
the rating torque is 4500 kNm. The value of actual
torque/rating torque is 0.81; and the value of additive
stress/max. stress is 0.32. So according to the earth
pressure and the torque and assessing by the Table 3,
the EPSR is about 0.65.
4.3
4.0
m 3.0
2.0
1.0
0.0
0.05
0.1
0.15
MPa
Pressure properties in cobble sand strata
The line 1 of Chengdu metro is in strata of cobble and

sand. The EPB machine is
6.28 m of Herrenknt. The
main parameters of the machine are shown in Table 4.
From the analysis hereinbefore, the EPSR is about
0.6 0.7 in the soft clay strata. And in most of the situations the EPB tunneling cant reach total earth balance
in working face. The values of EPSRs are usually controlled bigger than 0.5. The main balance pressure is
earth in this stratum. The regularity of the earth pressure is good. The regular modulus of earth pressure
reaches 0.98.
The tunnel line is in the strata of cobble and sand.
The cobble content is high and the permeability coefficient is high. The main indexes of strata <3 7> are
listed in Table 5.
The pressure data in the Figure 7 are collected in the
ring of 112 which is the mean value in the ring. And
the torque of cutter head is 3020 kNm in this ring. The
EPSR is about 0.4 from the Table 3 and Figure 7.
The regular modulus of earth pressure RMEP in the
chamber is 0.93 which is less than the value in clay. The
pressure gradient can reflect the muck unit weight. The
gradient value is the product of unit weight and lateral
Figure 7. Mean earth pressure in the pressure wall.
pressure coefficient. Due to the irregularity of earth

press in cobble and sand strata, its hard to calculate
the unit weight. However, it shows the unit weight is
bigger according to Figure 6 and Figure 7.
The data show that the EPSR is low. So the cutter
head plate will undertake bigger pressure. Low EPSR
is unfavorable for wear of cutters and cutter head. So
the EPSR should be increased by taking some measures. Such as soil conditioning should be taken to
increase the mucks plastic fluidity. Simultaneously,
the regular modulus of earth pressure can be increased.
4.4
Buffer ability of working chamber
The discharging ability of the screw conveyor is as

below:
347
where:- discharging efficiency, if there is no loss,

= 1. In common is about 0.9; D1 - inner diameter
of screw conveyor; D2 - outer diameter of center shaft;
P- Distance of paddle; n- Rotation speed.
According to formula (6) the maximal discharging
ability of screw conveyor is 1.2 m3 /min. The volume of
discharging is 6 m3 in 5 min. The cubage of the chamber is about 30 m3 .According to formula (5), the buffer
ability is 20% (6/30 = 20%) of the cubage of chamber
(shown in formula 7).
According to calculation, 30% loss of working chamber volume can cause big ground settlement. The
bigger buffer proportion can cause more ground loss
and instability of working face. If the proportion is
over 30%, it can be deemed that the buffer ability of
the chamber is not enough. If it is below 30% it can
be deemed that the chamber has the ability of earth
pressure buffer for improper operation.
5
CONCLUSIONS
From the analysis and study of the theory and measured data, the main conclusions are as below:
1 The classical earth pressure theory has limited and
premises.
2 The pressure in the pressure wall is divided into two
parts: additive pressure of working face and lateral
pressure of mucks.
3 The conceptions of earth pressure supporting ratio,
regular modulus of working chamber and buffer
ability of chamber are put forward to research the
pressure properties.
4 Clay stratum and cobble sand stratum are taken for
examples to study.
5 The research shows that the actual pressure cant

reach the situation of classical pressure.
6 EPSR and regular modulus in clay are both higher
than in cobble and sand. So the invalid wear is
higher in cobble and sand strata.
REFERENCES
Abdul-Hamid Soubral 2000. Three dimensional face stability
analysis of shallow circular tunnels. International Conference on Geotechnical and Geological Engineering, 2000.
Australia: Melbourne.
Anagnostou G & Kovari K. 1996. Face stability conditions earth pressure balanced shields. Tunneling and
Underground Space Technology11(2):165173.
Herrenknecht, M. & Rehm, U. 2003. Earth Pressure Balanced
Shield Technology. Internal Lecture in Colorado School
of Mine, USA.
Hisatake M. & Eto T Murakami T. 1995. Stability and failure
mechanisms of a tunnel face with a shallow depth. Proceedings of the 8th Congress of the International Society
for Rock Mechanics. Japan: Fujii.
Johnson, R.A. & Bhattacharyya, G.K. 1992. Statistics: principles and methods. New York : Wiley.
Qin, J.S. 2005. study on mechanism of the face deformation
and failure in shield tunneling. P.H.D thesis, University of
Hohai.China: Nanjing.
Shikibu, N. 1995. Stability of face during shield tunneling
A survey of Japanese shield tunneling 1995. In: Fujita,
Kusakabe. Underground Construction in Soft Ground.
Rotterdam: Balkema.
Stallebrass, S. E., Grant, R.J. & Taylor, R.N. 1996. A Finite
element study of ground movements measured in centrifuge model tests of tunnels. Geotechnical Aspects of
Underground Construction in Soft Ground, 1996. Rotterdam: Balkema.
Wassmer, L., Treceno, O. & ANdreossi, E. 2001. Tunnel boring machine(TBM)Application in Soft Ground Condition.
I.M.I.A. Meeting 2001, Australia: Sydney.
348
Backfill grouting research at Groene Hart Tunnel

A.M. Talmon
Deltares (Delft Hydraulics) & Delft University of Technology, Delft, The Netherlands
A. Bezuijen
Deltares (GeoDelft) & Delft University of Technology, Delft, The Netherlands
ABSTRACT: Measurements on back-fill grouting at Groene Hart Tunnel (GHT) are described and analysed.
The research utilises TBM data, monitored grout pressures and measured rheological properties of tunnelling
fluids. Grout injection strategy and vertical grout pressure distribution behind the TBM were found to agree
with established theory, like found in other tunnelling projects. However, on some occasions different smaller
vertical grout pressure gradients were found, which might have been caused by an incidental upward movement
of the tunnel. It is shown that the consolidation of grout determines the decay of grout pressures. The associated
net hydraulic resistance of surrounding soil is substantially higher than according to radial flow theory. This
higher resistance is attributed to bentonite originating from the face of the TBM, having invaded the grout-soil
interface.
INTRODUCTION
The Groene Hart Tunnel (GHT) in the Netherlands is

part of a high speed railway line between Amsterdam
and Brussels. The location of the tunnel is about 20
kilometers South-West of Amsterdam. The length of
the tunnel is 7.16 km. The tunnel is situated at a depth
of about 30 meters. Hydro-geotechnical conditions are
described by Aime et al. 2004. This double track tunnel
has a diameter of 14.5 meters. Aristaghes et al. 2002
reported on the design of the tunnel.
Grouting research at GHT was conducted as a
cooperation between Delft Institutes, HSL project
organisation and COB-research. The study focused on
the interaction of soil, grout and tunnel construction
at the injection of grout. Three specific aims were:

1) Analysis of grout pressures measured at GHT.
2) Quantification of the consolidation of grout and its
influence on grout pressures. 3) Modelling of the grout
pressure distribution; specifically from grout injection
at the TBM to the buoyancy dominated region further
behind the TBM. Novel conditions at the time were
the impressive diameter of the tunnel: 14.5 m, in contrast to the 10 m diameter tunnels monitored up to then,
the critical hydro-geotechnical soft soil conditions and
a different type of grout. The geotechnical profile is
shown in Fig.1.
Laboratory characterisations were conducted on
the grout: vane testing for rheological properties,
consolidation experiments, and tests to determine to
Figure 1. Geotechnical profile Groene Hart Tunnel and measurement locations: Instrumented plot no 1 at ring 344
(29/03/2002), COB-passage at ring 2117 (03/06/2003) and Noordplaspolder at ring 3241 (05/11/2003).
349
shear stress [kPa]
on the contrary, is nearly constant in time. The grout

was fine grained, though no sieve tests were conducted
for quantification. Similar tests were conducted on
grout collected in December 2003.Also the shear modulus of grout was determined. This property is needed
to calculate the transition to elastic behaviour (=input
to DCLong: Talmon & Bezuijen 2005).
The peak shear strength, instead of the remoulded
strength, is here the relevant parameter, since through
continued consolidation the failure surfaces are closed
continuously, and it is here that the peak strength has
to be surmounted continuously.
peak strength
remoulded strength
peak strength (Dec 03)
4
3
2
1
0
0
10
20
30
40
50
hours since production [hh]
Figure 2. Peak shear strength and remoulded strength of

GHT tail void grout, determined by vane testing.
what degree of grout consolidation the grout pressure transducers used in the lining will give reliable
results. Also the rheological and consolidation properties of bentonite that could surround the grout were
measured.
The construction of Groene Hart Tunnel was monitored on a number of locations. Three locations were
available for Delft Institutes research: Instrumented
plot no 1, a so-called COB-passage and a passage
in Noordplaspolder, see Figure 1.
In each of these locations grout pressures on the
tunnel lining were measured. The COB-passage had
additional instrumentation: pore water pressure measurements, axial and tangential strain measurement in
the tunnel lining, convergence measurements, vertical position measurements of the tunnel lining and tilt
measurements of the tunnel lining. This paper focuses
on grout pressures.
2
2.1
PROPERTIES OF TUNNELING GROUT

Rheology of grout
Yield stresses are governing grout pressures behind

a TBM. Shirlaw et al. (2004) describe the ingredients of the cement-less grout employed at Groene Hart
Tunnel (GHT): sand, flyash, lime and chemical additives.They also address some of its physical properties.
Fresh grout samples were taken (27/03/2002) from a
supply container located in the TBM. The grout was
tested in Delft. The density of the grout is 1850 kg/m3 ,
the water content is 0.201.
Peak shear strength and remoulded shear strength
were measured with a vane apparatus as a function
of time. Figure 2 shows the results. The peak shear
strength of fresh grout is about 1 to 1.5 kPa. The tests
also show that peak shear stresses increase slowly in
time. This is caused by puzzolanic reaction of the flyash present in the grout. The remoulded shear strength,
2.2 Consolidation of grout

In porous soil conditions elevated grout pressures,
compared to the pore water pressure at the same location, lead to consolidation of grout, or grout bleeding.
The consolidation of grout, in turn, influences the
effective stress distribution some distance behind the
TBM. Consolidation starts at the grout-soil interface,
and proceeds slowly into the grout layer that fills the
tail void (this layer is called a grout cake). Consolidation properties of GHT grout were determined in
Delft by means of element tests such as described by
Bezuijen & Talmon (2003). Permeability and porosity change of the grout are the governing parameters.
Relevant results to the calculation of grout consolidation in the tail void are: permeability grout cake
k = 2.4*108 [m/s], porosity fresh grout ni = 0.31 and
porosity of the consolidated grout cake ne = 0.24.
3
LABORATORY TESTING OF GROUT

PRESSURE SENSORS
Pressure sensors (SenSym/ICT 19C 100 P) have been

mounted in the tunnel lining segments. A cavity
(D = 82 mm, h = 23 mm) filled with tail brush grease
separated the sensor membrane from the grout (this
construction is comparable to that at Sophia Rail
Tunnel, which is described by Bezuijen et al. 2004).
Since at the GHT beam action of the tunnel lining
was the primary subject of the COB-research, it had to
be sure that the grout pressure data was reliable, specifically under consolidated/hardened conditions of the
grout. The functioning of these grout pressure sensors
was tested in the laboratory, Bezuijen (2004). A sketch
of the test cell is shown in Figure 3. A pressure sensor was placed in a grease filled cavity position, as at
GHT.
The grout in the test cell was loaded by increasing
the air pressure in the top of the cell. The resulting
grout pressure was the same as in the field.
Up to about one hour, this sensor adequately measured externally imposed variations of total pressure,
see Figure 4. After that the pressure readings were in
350
pressure [kPa]
valve
0
50
100
150
200
250
300
350
400
pressurized
air
vertical position[m]
-9
geotextile
grout
0.2m
grout press.
grease gauge
water
collection
plate
-6
-3
0
3
6
28 m behind TBM
pore water
slurry front
total press.
gauge
Figure 5. Pressures measured on tunnel lining at 28 m

behind TBM.
pore pressure gauge
Figure 3. Laboratory testing of grout pressure sensor.
GROUT PRESSURES MEASURED AT GHT
4.1 Measured grout pressures at Instrumented

plot no.1
350
20
displacement
15
air press.
250
grout press.
10
200
GD tot. press
150
pore press.
100
-5
50
-10
-15
-50
displacement (mm)
Pressure (kPa)
300
-20
0
time (hours)
Figure 4. Example of pressure sensor response.
between the imposed total pressure and the pore-water

pressure.
During this process another flush mounted pressure
sensor with very stiff membrane, continued to measure
the total stresses. It is concluded that after one hour
consolidated grout reached the pressure sensor. Subsequently grout-arching might have occurred in front
of the GHT pressure sensor when loaded dynamically.
As a consequence, pressure readings at GHT have to
be interpreted with care. It should however be mentioned that these tests might have been too stringent;
the imposed pressure variations are not likely to occur
in the field. Instead there is a slow decay of grout
pressure with time, given arching less chance to occur.
Furthermore the sensor is located at the impermeable
lining. In Section 6 it will be shown that in a tunnelling
process it can take more than 25 hours before the grout
close to the lining is consolidated
Twelve grout pressure sensors were utilised in Instrumented plot no 1. The time-series were relatively
short: the data covered at max 3 rings behind the TBM.
The wireless FM-data transfer of the sensors suffered
interruptions and malfunctioning.
The grout was injected through six injection pipes.
The pressure drop over the pipes was about 3 bar.
This pressure drop is equivalent to a wall shear stress
of 1 kPa. This is significantly higher than found for
cementious grouts such as at Sophia Rail Tunnel and
Botlek Rail Tunnel, where about 0.2 kPa was found
(Talmon et al. 2001).
Pump-stroke counts showed that about 60% of the
grout was injected through the three injection ports in
the upper half of the TBM. Thus a net downstream of
grout is created at the back of the TBM.
Despite the sparse set of tail void pressure data,
a picture emerged that the vertical grout pressure
gradient behind the TBM (10 kPa/m) could have
been rather small in comparison to earlier tunnelling
projects such as Sophia Rail Tunnel (Bezuijen et al.
2004).
4.2 Measured grout pressures at COB passage
At COB passage, tail void grout pressures were measured by a total of 32 pressure sensors. Ring 2117
and 2118 were each circumferentially equipped with
10 pressure sensors. The remaining 16 sensors were
equally distributed over ring 2119, 2120 and 2121.
The grout pressure profile, 28 m behind the TBM,
is shown in Figure 5. At this point it approaches the
hydrostatic pressure distribution. Immediately behind
the TBM, measured pressures are about 100 kPa above
pore water pressure. Only two sensors located in the
351
Figure 6. Mean grout pressure of ring 2117, 2118 and 2120.
Figure 7. Vertical grout pressure gradients measured at ring

2117, 2118 and 2120. Between 16 and 25 metres electronics
did not function correctly.
lower half of the tunnel, strangely, showed pressures

equal to pore water pressure.
Average grout pressures measured at ring 2117,
2118 and 2120 are shown in Figure 6. It shows that
grout pressures decay with distance from the TBM,
in correspondence with the fundamental processes
described in Talmon & Bezuijen (2005).
The pressure distribution over the circumference
of the tunnel lining is determined by specific weight
of the grout and flow resistance when the grout is distributed around the lining. The variation of the vertical
grout pressure gradient with distance behind the TBM
is shown in Figure 7. This figure shows that at upon
exiting of ring 2117 this vertical pressure gradient is
rather small: 6 kPa/m. This implies that there has been
a down-flow of grout with a high flow resistance.
Measured vertical displacements of the tunnel lining (measured by a water levelling system measuring
vertical displacements between ring 2117, 2118 and
ring 2057) indicate that the tunnel lining had moved
to a high position before the pressure sensors came
into the grout. Before exiting of Ring 2117, the lining
had already moved upward some 6 cm, see Figure 8.
Ring 2057, located 120 m behind the TBM served as
a reference.
Figure 8. Vertical position of ring 2117 measured by a liquid

leveling system and tilt of ring 2117 measured when the grout
sensors first came into the grout (3-6-2003 06:00).
Ring 2117 and 2118 were also equiped with tilt sensors that were attached to tunnel ring segments situated
at tunnel axis height. The time trace of the tilt meters
varied in concert with vertical positions measured with
the liquid levelling system. Near the end of drilling for
ring 2120 (3-6-2003 12:00), the liquid levelling system had to be dismantled, the tilt measuring system
continued measuring though. The tilt-measuring system showed that after the drilling of ring 2120, the tilt
had dropped 0.04 degree, which according to Figure
8 corresponds to a vertical downward movement of
about 6 centimetres since ring 2117 and 2118 came
into the grout.
Overviewing the data, it is concluded that when
ring 2117 and 2118 came into the grout the tunnel
lining was at a temporally high position. The course
of tilt meter data, vertical pressure gradient and vertical position correspond qualitatively. A representative
vertical grout pressure gradient at the back of the TBM
is 12 kPa/m.
4.3 Measured grout pressures at Noordplaspolder
The data from Noordplaspolder produced a more
familiar picture: vertical grout pressure gradients of
about 15 kPa/m immediately behind the TBM and a
vertical gradient of about 8 kPa/m at 17 meters behind
the TBM, see Figure 9.
5 ANALYSIS OF GROUT PRESSURES BEHIND
THE TBM
A finite difference model for the calculation of grout
pressures immediately behind a TBM was first presented by Talmon et al. (2001). The model calculates
the distribution of grout issued from the injection
ports, and consequently predicts the pressure distribution directly behind the TBM. A tail void of constant
thickness is assumed and the computational domain
stretches for about 5 meters behind the TBM.
352
8
behind TBM (drilling)
8 rings behindTBM
height [m]
water
grout static
0
-2
-4
-6
-8
0
100
200
300
400
500
600
grout pressure [kPa]
Figure 9. Vertical grout pressure profiles measured at ring

3241 of Noordplaspolder.
Figure 10. Grout pressures measured up to 1 m behind the

TBM at Instrumented plot no 1 and calculated pressures
(model DCgrout). s = axial coordinate. Two-sided friction
grout layer: y = 1.5 kPa, 60% of grout is injected through
injection ports in the upper half of the TBM.
When the results of the Instrumented plot no 1

came available a calculation of the grout pressure distribution immediately behind the TBM was made, see
Figure 10. Inputs were the vane test results of March
2002 and the measured distribution of flow rates over
the six grout-injection pipes.
Only if two-sided friction of the grout layer is
assumed, like in the original model, a fair correspondence with the measurements is achieved. A pressure
difference of nearly 150 kPa is calculated between
the bottom and the crest of the tunnel. This pressure
difference is equivalent to a vertical grout pressure
gradient of 10 kPa/m. For reference, self-weight of the
grout would produce a vertical pressure gradient of
18.1 kPa/m. For 300 kPa at the crest, this would lead
to a pressure of 562 kPa at the bottom of the tunnel.
When first data of the COB-passage came available it was found that the distribution of flow rates
over the injection ports was quite comparable: 55%
of the grout was injected through the three upper

ports. The grout pressure data (ring 2117) showed a
small vertical grout pressure gradient of about 4 a
6 kPa/m, see Figure 7. This could not be explained by
means of the DCgrout model. The measured vertical
pressure gradient more is compatible with a situation where all of the grout is forced downwards. In
that case the vertical pressure gradient is given by:
18.1-2*y /h = 3.1 kPa/m.
At ring 2118 and 2120 a vertical grout pressure gradient of about 12 kPa/m is measured behind the TBM.
Here the results correspond more with a situation
where grout distributes in both the up- as downward
direction from the injection ports.
The pace of the construction cycle at COB-Passage
was slowed down by installation of instrumentation.
The measured pressure distribution at Noordplaspolder is more in line with earlier experience at
other tunnels: f.i. Sophia Rail Tunnel (Bezuijen et al
2004).
Given the rheological properties of grout, the measured vertical grout pressure gradients can only be
achieved when there is friction on both sides of the
grout layer: friction at the soil-grout interface and
friction at the tunnel-grout interface. Given the high
frictional resistance measured over the supply lines, it
seems justified to conclude that this grout is providing significant friction with the concrete tunnel lining
(contrary to cementious grout employed in previous
projects).
It has been found that the pressure distribution
immediately behind the TBM varies strongly between
measurement locations. A likely candidate that might
have affected the pressure distribution is vertical
movement of the tunnel lining. The higher yield stress
of the grout used, compared to the grout in the earlier tunnels, has as a consequence that the movements
of the lining have a larger influence on the pressure
distribution around the lining.
6
CALCULATION OF THE CONSOLIDATION

OF GROUT
Pressurized grout situated against a permeable surface

is subject to fluid loss. Two different situations can be
discerned: conditions during stand still of the TBM
and conditions during drilling of the TBM.
Under stand-still conditions no fresh grout is supplied and grout pressures drop slowly. The pressurised
soil will unload. Filtration properties of grout were
measured in the laboratory, and a model was made
for grout consolidation during stand still (see Bezuijen and Talmon 2003). This situation is verifiable on
measured pressure decay during standstill. Involved
properties are permeability of grout, porosity changes
of the grout, hydraulic resistance of surrounding soil
and elasticity of surrounding soil.
353
of the tunnel occurs by consolidation of the grout. This

conclusion was also reached for the Sophia Rail Tunnel
(Bezuijen et al. 2004), where cementious grout was
used.
Figure 11. Skematisation of consolidation of grout. Elevated grout pressures are represented by the force F, the
hydraulic resistance of pore water flow is represented by R.
Figure 12. Numerical simulation of grout pressure decay

behind the TBM at COB-passage compared with measurement: calculated pressure decay and calculated thickness of
the grout cake, at R = 2 107 [s].
Under drilling conditions fresh grout is supplied

and the grout pressure is constant. The surrounding soil will be under a constant pressure, and only
permeability and hydraulic resistance are relevant.
The consolidation process is schematised in Figure
11. The hydraulic resistance R was back-calculated
from observed grout pressure decay. Measured pressure decay and calculated grout pressures match well,
see Figure 12. This back-calculated hydraulic resistance is however substantially higher than that of
surrounding soil. A typical value for the hydraulic
resistance of radial pore water flow into the surrounding soil is R = 65000 [s], Bezuijen (2005). It is
hypothesised that bentonite slurry from the face of the
TBM has invaded the grout-soil interface.
The consolidation theory allows the calculation of
the development of the grout cake along the grout-soil
interface, see Figure 12. The thickness of this filter
cake is important to the modelling of the grout layer as
an interface between tunnel lining and soil: it increases
the integral stiffness of the grout layer. The foundation
CONCLUSIONS
The fine grained grout at GHT has a yield stress of

about 1 a 1.5 kPa and, due to absence of cement, the
yield stress of this grout increases only slowly with
time. It was shown that consolidation of grout determines the decay of grout pressures over at least five
tunnel rings behind the TBM. The developing grout
cake increases the integral stiffness of the grout layer
and hence the tunnel is founded.
The associated net hydraulic resistance of surrounding soil is substantially higher than according to radial
flow theory. This higher resistance is attributed to bentonite originating from the face of the TBM, having
invaded the grout-soil interface.
Given the high frictional resistance measured over
the supply lines and the small vertical grout pressure
gradients behind the TBM, it is concluded that this
grout is providing significant friction with the concrete
tunnel lining (contrary to cementious grout employed
in previous projects).
For a number of reasons the pressure readings at
GHT were less reliable than at other tunnels. The GHT
data is nonetheless valuable because it shows the general applicability of the results from earlier research at
Sophia Rail Tunnel, it is providing data on the performance of cement-less grout with high yield stress and
it shows a marked influence of vertical movements of
the tunnel lining on the grout pressure distribution.
ACKNOWLEDGEMENT
The work was conducted as a co-operation between
Delft Cluster, HSL-South Organisation and Centre
for Underground Construction. The authors gratefully acknowledge HSL South Organisation, and their
treasurer the Dutch Public Works Department, and
Centre for Underground Construction for permission
to publish.
REFERENCES
Aime, R., Aristaghes, P., Autuori, P. & Minec, S. 2004. 15 m
Diameter Tunneling under Netherlands Polders, Proc.
Underground Space for Sustainable Urban Development
(ITA Singapore), Elsevier.
Bezuijen, A. 2004. Calibration grout pressure sensor
GHT/Results of measurements (in Dutch: Ijking groutdrukopnemers GHT/Resultaten metingen), COB report
F512-O-04-127.
354
Bezuijen, A. 2004. Vin tests grout GHT/Results of measurements (in Dutch: Vinproeven op grout GHT/Resultaten
metingen), GeoDelft 403050/0008.
Bezuijen, A. 2005. Analysis of pore pressures close to
the tunnel; influence of grout flow, (in Dutch: Analyse
waterspanningen naast tunnel; invloed groutstroming op
waterspanning), COB report F512-05-04.
Bezuijen, A. & Talmon, A.M. 2003. Grout the foundation
of a bored tunnel, 2003, Proc. int. conf. on Foundations
Innovations, Observations, Design & Practice (ICOF
2003 Dundee), Thomas Telford, London.
Field measurements on grout pressures during tunnelling
of the Sophia Rail Tunnel, J. of Soils and Foundations of
the Japanese Geotechnical Society,Vol.44, no.1, pp3948.
Shirlaw, J.N., Richards, D.P., Ramond, P. & Longchamp, P.

2004. Recent experience in automatic tail void grouting
with soft ground tunnel boring machines, in proc. Underground Space for Sustainable Urban Development (ITA
Singapore), Elsevier.
Talmon, A.M. & Bezuijen, A. 2005. Grouting the tail void
of bored tunnels: the role of hardening and consolidation of grouts, 5th int. Symp. Geotechnical Aspects of
Underground Construction in Soft Ground, 1517 June,
ISSMGE-TC28, Amsterdam, The Netherlands.
Talmon, A.M., Bezuijen, A., Aanen, L. & van der Zon, W.H.
2001. Grout pressures around a tunnel lining, proc.
IS-Kyoto conference on Modern tunneling Science and
Technology, pp817822.
355
Longitudinal tube bending due to grout pressures

A.M. Talmon
Delft Hydraulics & Delft University of Technology, Delft, The Netherlands
A. Bezuijen
GeoDelft & Delft University of Technology, Delft, The Netherlands
F.J.M. Hoefsloot
Fugro Ingenieursbureaux B.V., Leidschendam, The Netherlands
ABSTRACT: A mathematical relation is described between the distribution of grout over injection ports,
rheological properties and vertical grout pressure gradient immediately behind a TBM. This vertical grout
pressure gradient is an important loading parameter in beam action calculation of tunnel linings. Longitudinal
moments in the lining are of importance in tunneling especially for the design of the lining. These moments are
determined by the uneven distribution of jack forces from the TBM but also by the buoyancy forces that are
exerted by the grout that is still in the liquid phase close behind the TBM.
INTRODUCTION
Earlier work (Bezuijen & Talmon 2005) has shown

that the grout properties have a large influence on
the longitudinal moments and that it is even possible that these moments prevent an adequate tunnelling
process.
This earlier work did not take into account that
the tunnel is stage constructed and that a specific
beam calculation method has to be used to take
that into account. An analytical staged beam calculation will be used in combination with measured
and calculated grout pressure distribution to calculate
longitudinal bending moments in the lining. The calculation method itself will be described in another paper
to this conference (Hoefsloot, 2008).
The vertical grout pressure distribution immediately behind a TBM is governing to a large extent
the stresses in the tunnel lining. This vertical grout
pressure distribution is determined by the injection
strategy and rheological properties of the grout. A
finite difference model was published by Talmon et al.
(2001) for the calculation of the grout pressure distribution up to a distance of a few tunnel rings behind
a TBM. Using this model it was learned that there
might be a simple relation between the distribution of
grout injection rates over the injection ports and the
vertical grout pressure gradient immediately behind
a TBM.
2.1
SIMPLE MODEL FOR VERTICAL

GROUT PRESSURE DISTRIBUTION
AT TBM
Injection conditions
Many TBMs are equipped with six injection ports.

Typically the flow rate through the ports located in the
upper half of the TBM is larger than in the lower half.
This way, the grout is forced to flow downwards, partially. As a consequence, the vertical grout pressure
gradient will be lower than the static grout pressure
gradient. Such a grout injection pushes the tunnel
downwards. A definition sketch of grout distribution
is given in Figure 1. Data on typical grout injection at
Groene Hart Tunnel is given in Table 1.
The outer diameter of the Groene Hart Tunnel lining
is 14.50 m. The theoretical thickness of the grout layer
is 0.18 m.
2.2
Mathematical modeling of tangential

grout flow
Each injection port has its own supply area. Grout flow
velocities in the grouting annulus are small. Given the
Bingham-like rheological character of grout, only its
yield-stress is relevant. As a consequence, for both
clock and anti-clock wise direction, grout pressures
decay linearly from injection ports. Grout pressure
357
Figure 1. Definition sketch tangential grout distribution

behind a TBM, grout supply area and the z-coordinate of
the centre of gravity of grout injection.
Table 1. Measured distribution of grout flow rates over the

injection ports at Groene Hart Tunnel.
Injection
point
position
[degrees]
vertical
position
[m]
contribution to
total grout
injection
1
2
3
4
5
6
total
8.4
65.2
122
180.9
237.7
294.5
7.32
3.10
3.92
7.40
3.95
3.07
0.2
0.178
0.156
0.133
0.156
0.177
1
Figure 2. Calculated grout pressure distribution from grout

injection. Injection ports are indicated. A linear approximation of calculated grout pressure distribution is indicated.
The theoretical static grout pressure distribution is shown for
comparison. Grouting parameters are according to Table 1,
physical parameters are according to Table 2.
Table 2. Input-parameters for the calculation of

grout pressure distribution at Groene Hart Tunnel.
Dtunnel [m]
hannulus [m]
grout [kg/m3 ]
y [Pa]
Pcrest [Pa]
14.5
0.18
1850
1500
200000
2.3 Simplification of vertical pressure gradient

calculation
decay halts upon convergence with grout from adjacent
injection ports. The width of supply area is dictated by
the relative injection rates of associated injection ports.
Considering the width of supply area, and the fact
that no discontinuities in grout pressure are allowed,
the precise locations of the stagnation points between
supply area are calculated. Next, with yield stresses
on both sides of the grout layer, the grout pressure
distribution is calculated mathematically. The location
of injection ports and the definition of supply area is
given in Figure 1.
An example of the calculation results for Groene
HartTunnel is given in Figure 2. In this calculation both
sides of the grout layer experience tangential shear
stresses equal to the yield stress of fresh grout. Physical
properties are listed in Table 2.
In reality there are no sharp discontinuities at the
confluence of the supply area. These discontinuities
are an artifact inherent to the employed schematization. The calculated discontinuities are acceptable
because locally clock-wise and anti-clock-wise flow
compensate.
The loading force (F) of grout on a tunnel lining, per

unit length, is given by:
The average vertical grout pressure gradient might

be determined from linear approximations as shown
in Figure 2. This is however not the most straightforward method. The mathematical model presented
in Section 3 gives grout pressures at each circumferential position. By integration of calculated grout
pressures the loading force (F) is calculated. By virtue
of the above equation the corresponding vertical grout
pressure gradient is determined next.
A number of calculations was conducted for different injection strategies, where the grout distribution
between the upper and lower part of the TBM varied. Upon inspection of the results it was found that
the centre of gravity of grout injection is a decisive
358
Figure 4. Back-side of the TBM showing tail-grease

brushes and play (=jeu) between tunnel lining and back-side
of the TBM-shield.
Figure 3. Calculated dimensionless contribution of
grout-injection to vertical grout pressure gradient. Location
of injection ports according to Table 1.
parameter. The z-coordinate of the centre of gravity is

calculated by:
with: Qi = grout-flow-rate of injection port (i),

Qtotal = total grout-flow-rate of all injection ports,
zi = vertical coordinate injection port (i),
zg = z-coordinate centre of gravity
Definitions for the centre of gravity are shown in
Figure 1.
Calculation results for vertical grout pressure gradient from grout injection (without static grout pressure)
are shown, in Figure 3. Quasi independent of the precise distribution, a linear relation is found between
the centre of gravity of grout injection and vertical
grout pressure gradient. With six-injection ports, without grout overlapping adjacent injection ports, the
maximum vertical position of the centre of gravity is
zg (max) = R/3, and the influence of grout injection
is maximal: dp/dz = 8/y /h. In that case the injection
port at the crest of the tunnel supplies 1/3 of total grout
injection.
Adding the static grout pressure to the results of
Figure 3 leads to the following design-formula for
vertical pressure gradient immediately behind a TBM:
For the grouting conditions at Groene Hart Tunnel,

given in Table 1, the vertical position of the centre of gravity is at zg = 0.047 R. According to the
simplified theory presented above, a vertical grout
pressure gradient of 15.6 kPa/m would be expected at
Groene Hart Tunnel. Measured vertical pressure gradients measured at Groene Hart Tunnel were: (5 kPa/m
at ring 2117, 12 kPa/m at ring 2118 and 17 kPa/m at
ring 3241, see Talmon & Bezuijen 2008). With the

method described in Section 2.2 and 2.3 the vertical
grout pressure gradient behind a TBM is calculated
as a function of injection strategy and grout properties. This method is a good alternative for the DCgrout
model. The advantage is that no numerical simulation
is needed. The disadvantage however, is that it only
presents the gradient directly after the tunnel. In the
calculations that follow we assume that in 8 m this gradient decrease down to 9.3 kPa/m as will be explained
later.
2.4 Grout pressures in eccentric tail void

The grout pressure calculation model of Section 2.2
assumes a concentric position of the tunnel lining in
the soil cavity, as does the DCgrout model of Talmon
et. al. (2001). At Groene Hart Tunnel however some
7 cm upward movement was measured by a water leveling system just before the pressure sensors came into
the grout, Talmon & Bezuijen (2008). It is unknown
if, or how, surrounding soil moved. A movement of
around 2 cm was found in 2-D calculations by Bezuijen
& Bakker (2008) for a tunnel of 10 m diameter. However, this result will depend to a large extend on the soil
parameters and they were not obtained for this location. The results of these 2-D calculations do indicate
that a soil movement of some centimeters is the order
that can be expected. From a geometric point of view,
the maximum relative movement between the tunnel
lining and the rear of the TBM is at max 9 cm before
the tunnel lining touches the tail-end of the TBM, see
Figure 4.
It is concluded from Figure 4 that the maximum
eccentricity between the tunnel lining and soil cavity will be about 45 mm. The existing models do not
necessarily have to be modified. An adaptation of
inputted grout flow rates is considered a workable
alternative. For 45 mm upward eccentricity of the tunnel lining at GHT, the apparent centre of gravity has
been calculated at zg = 0.22 R.
359
Figure 5. Loading diagram tunnel lining Groene Hart

Tunnel: loading per unit length.
Figure 3 shows that for a relative centre of gravity of zg /R = 0.22, grout injection produces a vertical
pressure gradient of dp/dz = 1.6 y /h = 13.6 kPa/m
for a yield stress of y = 1.5 kPa and a tail void
thickness of h = 0.18 m. Addition of the static grout
pressure gradient (at = 1850 kg/m3 ) gives for the
calculated vertical pressure gradient behind the TBM:
dp/dz = 18.1 13.6 = 4.5 kPa/m. This is close to the
measured initial vertical gradient of 4 a 6 kPa/m at
ring 2117.
2.5
Different positions of injection ports
The present application applies to the layout of the

Groene Hart Tunnel in the Netherlands. Similar calculations should be conducted for other positions of the
grout injection ports, for instance for a layout such at
Botlek Rail tunnel, where there is no injection port at
the crest. Expected is that the result (Figure 3) will not
differ much.
Figure 6. Bending moment exerted by TBM jacks during

drilling phase.
Figure 7. Measured bending moment in tunnel lining (computed from strain-gauges precasted in tunnel segments).
BEAM ACTION CALCULATION
3.1 Tunnelling conditions

The (local) uniform loads in the tunnel are sketched in
Figure 5.
The forces exerted by the main jacks of the TBM
are calculated from measured hydraulic oil pressures
and piston/barrel construction of the hydraulic cylinders. During exit of the first rings of the COB-passage,
ring 2117, the point of application of the axial force
was 1.5 m below the TBM-axis: the bending moment
exerted by the TBM jacks was 79 MNm.
The bending moment exerted by the TBM jacks was
found to be virtually identical for each ring, with the
exception of those first rings of the COB-passage, see
Figure 6. A representative value is 65 MNm. Assuming that the other external influences (transverse force
from TBM, vertical grout pressure gradient behind the
TBM and longitudinal uniformity of soil) remain constant, a time-distance transformation of axial strains
measured in tunnel ring 2117 produces the bending
moment-curve of the tunnel lining shown in Figure 7.
Figure 8. Average injection pressure of six grout injection

lines and the vertical position of the centre of gravity of grout
injection.
At the beginning of the COB-passage the variation

in bending moment exerted by TBM jacks is precisely
synchronous with the bending moment calculated
from the strain gauges in tunnel ring 2117.
Figure 8 shows the course of grout injection pressure (average of all six injection lines) measured 4 m
before the outlet of the injection pipes. Also the centre
of gravity of grout injection, calculated by Equation 2,
360
Figure 9. Average injection pressure and vertical pressure

gradient of grout injection lines, measured at the inlet of the
injection pipes.
is shown. Grout injection pressures vary between 4.5

bar and 7.5 bar. The vertical position of the centre of
gravity of grout injection varies between 0 and 0.2R
above tunnel-axis.
For a concentric configuration of tunnel lining and
soil cavity, at a grout yield stress of 1.5 kPa and twosided friction of the grout layer, this corresponds to
a variation of vertical grout pressure gradient at the
TBM between 7.1 < dp/dz < 18.5 kPa/m.
Figure 9 shows the vertical pressure gradient calculated from injection pressures. Basically this gradient
varies between 0 and 10 kPa/m over the COB-passage.
This gradient is lower than measured with the pressure
sensors on the lining. It could be that the pressure drop
over the four meters pipe length is higher in the upper
part of the TBM, because of higher grout flow rates
(typical grout flow velocity is 0.3 [m/s]).
TBM jacks may transfer transverse forces between
the tunnel lining and TBM. The magnitude of such
forces should be limited, because jacks are not
designed for that. This unknown transverse force is
however important to beam action of the tunnel lining,
and has to be determined.
In the tail void the segments are supported by forces
from the grout layer. Inside the TBM tunnel rings are
unsupported. At the Groene Hart Tunnel the typical
length of unsupported tunnel lining in the TBM is 4
meters.
Vertical support by fluid grout is determined from
grout pressures measured around the tunnel lining.
Upon exit vertical grout pressure gradients as low as
6 kPa/m occurred, that after 3 m progress stabilized to a
gradient of 12 kPa/m (Talmon & Bezuijen 2008). This
latter value is considered a realistic and representative
value for grout loading just behind the TBM.
3.2
Result beam action calculation
The staged beam action model described by Hoefsloot (2008) is used. Input-parameters are the above
Figure 10. Measured and calculated bending moments

compared (Groene Hart Tunnel, the Netherlands).
mentioned bending moment by TBM jacks, transverse

force by TBM, vertical grout pressure gradient behind
the TBM, loading diagram and unsupported length of
tunnel lining in the TBM.
Model-specific is the so called end of the fluid
grout zone. In the staged beam model, this is the location where the reaction-force from the soil is zero: new
elements are mathematically added in a stress-less condition. Here the associated vertical pressure gradient
across the tunnel lining is 9.3 kPa/m (weight of grout
layer has been subtracted from a 10 kPa/m hydrostatic
pressure). The measured position of 9.3 kPa/m is positioned about ten meters behind the TBM. The vertical
grout pressure gradient is assumed to decay linearly
from 12 kPa/m at the TBM to 9.3 kPa/m at the end of
the fluid grout zone. Consequently the external loading by fluid-grout connects smoothly to the reaction
force from the soil. A further second order condition
is that the axial gradient of vertical grout pressure gradient and reaction force from the soil are identical. An
end of fluid grout zone situated 8 m behind the TBM
satisfies these criterion. However, it is not clear how
general applicable the implicit boundary conditions
are. This will be subject to further research.
The model requires also information on structural
properties: bending stiffness of the lining and modulus of subgrade reaction of surrounding soil. These
have to comply with fundamental properties of segmented lining and surrounding soil, but also have to
comply with the vertical displacements that have been
measured (Talmon & Bezuijen 2008). Therefore lower
values are used then one would apply on pure theoretical grounds. The following values have been used:
EI = 3.2 109 [kN m2 ] and k = 7.3 104 [N/m2 ].
A modest downward force of 1.5 [MN] from the
TBM is needed to reach perfect agreement between
measured and calculated bending moment curve in the
tunnel lining, see Figure 10.
At ring 2117 and ring 2118 a total of four inclinometers meters have been attached to lining segments
361
The construction phase of a bored tunnel leads

to axial forces that remain in the tunnel lining after
completion (Blom 2002). The construction phase also
produces a remaining bending moment, as was found
at the Groene Hart Tunnel. The associated point of
application of this axial force is situated about 1.5
meter above the tunnel axis.
ACKNOWLEDGEMENT
Figure 11. Measured and calculated inclination of the

tunnel (Groene Hart Tunnel, the Netherlands).
situated at tunnel-axis-level. These four instruments

produced exactly the same results. Figure 11 shows the
results of one of these. It shows that after the TBM the
tunnel inclination is of the order 1:1000 (compared to
its final orientation) and that inclination decreases with
distance. The graph shows that there remains a discrepancy between inclination calculated by the beam
action model and measured inclination.
4
CONCLUSIONS
Grout injection:
The centre of gravity of grout injection, in combination with grout yield stress, is shown to determine the
vertical grout pressure gradient immediately behind a
TBM.
In case of an eccentric position of the tunnel lining,
the influence of eccentricity on grout pressures is also
quantifiable.
In future application of beam action models for tunnel linings, the model of Section 2 can be used to
calculate beam loading caused by the vertical grout
pressure gradient behind the TBM.
Beam action tunnel lining
The measured bending moment curve in the tunnel
lining is well reproduced by the beam action model.
The way how the transition from liquid grout with a
constant force on the tunnel to elastic behaviour is
incorporated in the model needs further study. Simulations have shown that the buoyancy forces that occur
at the location where the tunnel lining is floating
in liquid grout are essential to simulate the measured
bending moment.
Lower values for the bending stiffness of the tunnel lining and the modulus of subgrade reaction of
surrounding seem to apply than are expected on pure
theoretical grounds.
The work was conducted as a co-operation between

HSL-South Organisation and Centre for Underground
Construction. The authors gratefully acknowledge
HSL South Organisation, and their treasurer the Dutch
Public Works Department, and Centre for Underground Construction for permission to publish.
REFERENCES
around a TBM machine. Proc. 6st Int. Symposium on
Geotch. Aspects of Underground Construction in Soft
Ground, Shanghai
Bezuijen, A. & Talmon, A.M. 2005. Grout properties and
their influence on backfill grouting. Proc. 5th Int. symp.
Geotechnical aspects of underground construction in soft
ground, Amsterdam 1517 June.
Blom, C.B.M. 2002. Design philosophy of concrete linings
for tunnels in soft soil. PhD-thesis, Delft University Press.
Hoefsloot, F.J.M. & Bakker, K.J. 2002, Longitudinal effects
bored Hubertus tunnel in The Haque. 3rdInternational
Construction in Soft Ground IS Toulouse, 2325 october
2002.
Hoefsloot, F.J.M. 2008. Analytical solution longitudinal
behaviour tunnel lining, TC28 Shanghai.
Koek, A.J., Bakker, K.J. & Blom, C.B.M. 2006. Axial prestresses in the lining of a bored tunnel. 5thint. Symp.
Soft Ground, 1517 June, ISSMGE-TC28, Amsterdam,
The Netherlands
Talmon, A.M. & Bezuijen, A. 2005. Grouting the tail void
of bored tunnels: the role of hardening and consolidation of grouts. 5thint. Symp. Geotechnical Aspects of
Underground Construction in Soft Ground, 1517 June,
ISSMGE-TC28, Amsterdam, The Netherlands.
Talmon, A.M. & Bezuijen, A. 2008. Backfill grouting
research at Groene Hart Tunnel. Proc. 6st Int. Symposium
on Geotch. Aspects of Underground Construction in Soft
Ground, Shanghai.
Talmon, A.M., Bezuijen, A., Aanen, L. & van der Zon,
W.H. 2001. Grout pressures around a tunnel lining. Proc.
IS-Kyoto conference on Modern Tunneling Science and
Technology, pp. 817822.
Talmon, A.M., Bezuijen, A. & Hoefsloot, F.J.M. 2007. Evaluatie tweede fase Groene Hart tunnel: Ontwerpparameters bij krachtswerking, groutdruk en invloed beweging
tunnelbuis. COB-report, COB-F512-07-02.
362
Theme 3: Case histories
Tunnel face stability and settlement control using earth pressure balance
shield in cohesionless soil
A. Antiga
Soil S.r.l., Milano, Italy
M. Chiorboli
Metropolitana Milanese S.p.A., Milano, Italy
ABSTRACT: Soft ground tunneling, in urban areas are generally bored using pressurized TBM. Tunnel design
involves the need to revalue the classical concepts of deformation response to the excavation and the control
priorities, focalized on face and cavity, loose their importance; the evaluation of the settlements becomes the
main aspects to consider. The definition and the control of the face pressure are fundamental design steps.
It is important to give a proper importance to the face pressure relating to settlements control. The pressure
increase leads to higher safety factors of the face stability but it is not able to guarantee comparable settlements
decrease. We examines the essential parameters in relation to the use of the EPBS in two case histories in Milan:
line 1 extension to the new Fair and Passante rail infrastructure. The reference values of the cited parameters are
identified for cohesionless soil i.e. fluvio-glacial sand deposit and gravel.
INTRODUCTION
In recent years, the growing needs for transportation

infrastructure has lead to a sharp increase in the utilization of underground. During 1993 and 1994, an
EPBS with a diameter of 8.03 m was used in the area
of the Passante (underground railway line) in Milan.
This method, with a diameter of 6.56 m, was then used
in 2004 to extend the Milan subway to the new trade
fair of Rho. This document considers the main project
aspects relating to the use of the EPBS, with reference to the two cited projects in Milan and so to a
geotechnical context of cohesionless soil under high
water head. The reference values for the basic tunnel design parameters are defined; a reference value
for the parameter i (point of inflection in the transverse settlement trough) is identified. The document
also describes how, given the values necessary to
guarantee the face stability, the tunnel face pressure
would not influence the final settlements, and how this
parameter could be best defined using only calibration
during the work. The theoretical pressure values to use
can be determined using simple equilibrium ratios in
K0 conditions as initial starting values for tunnelling
operations.
MECHANIZED TUNNELLING DESIGN IN

URBAN ENVIRONMENT
Peck (Peck 1969) highlighted the basic points to consider when planning underground works in the case of
soft ground:
1 the stability of the tunnel with particular reference
to the stability of the tunnel face and of the zone
immediately behind;
2 the evaluation of the deformations at the face and
around the tunnel and, in the case of a tunnel with
a low depth, the resulting superficial subsidence;
3 the definition of stabilization actions to guarantee
compliance with conditions (1) and (2) and of the
lining structure to ensure the long-term stability of
the tunnel.
With the construction of tunnels in urban environment, the control of superficial subsidence and its
effects assumes primary importance with regard to
those items listed above. The conditions for stability
of the face and of the cavity and the acceptability of
respective deformations, with regard to the breaking
values, are conditions that are necessary but not sufficient to guarantee the validity of a given construction
365
Figure 1. Building subsidence: main parameters.
Figure 2. Transverse settlement trough.
system because they are not able to guarantee, a priori,

the acceptability of the subsidence. This statement is
particularly valid in the case of tunnelling systems with
shield machines and full-face pressurization for which
there is no direct control of the strain state of the face
and of the cavity during the construction phase, and for
which the most significant quantitative parameters are:
stabilization pressure at the face
volume of extracted soil
superficial subsidence
axis), (b) the distance i between the tunnel axis and

the point of inflection in the subsidence profile that
defines the size of the transverse settlement trough.
The vertical settlement at distance y on the tunnel
axis is given by:
The total volume of the subsidence trough Vs (for

the unit length for the tunnel) can be obtained from the
integration of (1) and the result is:
SUBSIDENCE EVALUATION
The process of design for tunnels in urban areas must

be based on the settlement value parameter.
The objectives of the settlement analysis should be
as follows:
definition of acceptable settlement values
study of the process that generates the subsidence
and the control methods
technical and economical optimization of the
project by defining the construction methods to use
and any required mitigation actions.
Burland and Wroth identified the parameters to consider when valuating subsidence (figure 1) (Burland
and Wroth 1974). Rankin subsequently focussed on
the deflection ratio DR = max /L and on the horizontal
deformation, he moreover showed that in engineering
practice, you can rarely rely on the complete data identified by Burland and Wroth (Rankin 1988). Similar
to a normal Gaussian probability function, the profile
of the transverse settlement trough is consolidated by
a mass of literature on the subject and is proven by
experience.
The curve that defines the transverse settlement
trough is characterized by two parameters: (a) the maximum settlement Smax (corresponding to the tunnel
The volume of the subsidence trough Vs is directly

dependent on the volume loss Vp (volume of soil that
is excavated in excess of the theoretical volume of the
tunnel).
In agreement with the proposal by OReilly and
New, for z > D (D = diameter of the tunnel), the
parameter i depends on the type of soil and the depth
of the tunnel and appears to be independent on the tunnel diameter and the tunnelling mode (OReilly and
New 1991); giving:
where z is the depth of the tunnel and K is a coefficient

that depends on the type of soil.
For cohesive soil, the value of K is generally
between 0.4 and 0.6. For sand and gravel, the values
for K are more dispersed, however the usual range is
between 0.25 and 0.45 (Mair 1997).
From the function (1) you can derive the slope and
curve of the transverse settlement trough, factors that
are relevant on the pre-existent structures.
To summarize, the value of the settlement is therefore dependent on i and on Vp through Vs .
366
Figure 3. EPB Shield _ Face stabilization pressure.
In the case of non-complex situations that can

be well planned following the hypothesis described
above, there are no significant improvements in the
reliability of the results after switching from an empirical Gaussian method to a more sophisticated numerical
model.
FACE STABILIZATION PRESSURE
The definition and the control of the face stabilization pressure represents one of the basic stages in the
constructive phase. The determination of this parameter in the planning phase, however, give rise to some
uncertainties: despite there being a large amount of
literature, there is not yet a generally agreed analysis
method and at the same time, there are no normative
references. Also, ignoring the uncertainties resulting
from the geotechnical characterization of the soil,
using the different approaches discussed in literature,
you obtain frequently significant differences in the
theoretical values to apply.
However, it is necessary to lay proper emphasis to
the face stabilization pressure in terms of the final
subsidence result.
On one hand, the increase in value of the pressure
applied to the face guarantees higher safety coefficients for the face stability; however, on the other hand,
a correlated reduction in settlements at the surface are
not absolutely guaranteed.
The definition of the stabilization pressure to apply
must achieve the objective of guaranteeing the stability
of the face while changing the in situ stress as little as
possible. In a fully-operational situation, you should
try to excavate in K0 condition, that is with a constant
advance rate.
A study conducted by AFTES identified how the
ideal advance condition, for minimizing the deformations and ensuring the face stability, is a balance
between the volume of material extracted and the theoretical tunnel volume and how with this condition
the pressure remains constant. In this condition, you
could advance with a constant volume and the pressure applied to the face would be equal to the earth
pressure at rest (i.e. K0 conditions); by checking the
two identified parameters (volume and pressure) you
would obtain the definition of the correct face stabilization pressure (AFTES 2001). As a result, during
construction phase, the research should concentrate on
the achievement of the balance between the volumes,
both theoretical and extracted, based on stabilization
pressure that is sufficiently cautious.
The machines used in recent years, mainly due to
the correct definition of the soil conditioning technologies, enable good control of pressures and, therefore,
of the stability of the tunnel face, as we will see in
the Milan examples. With the control of the pressure,
you can influence the situation ahead of the face of
the tunnel and this represents a condition necessary,
but not sufficient, to guarantee subsidence control. In
fact, to greatly increase the pressure would bring the
soil ahead of the cutting head in conditions nearer to
that of passive pressure with an increase in the total
pressure to apply and, as a result, with a series of effects
that bring about a reduction in the advance speed and
operating conditions that are less favourable and more
risky (for example, an increase in the required torque,
an increase in energy consumed, an increase in tool
wear, an increase of temperature in the tunnel working
chamber as a result of increased friction with the possibility of creating material blocks). The soil in front of
the face does not record any excess pressure applied to
the face in the successive phases in which it is subject
to the phenomena of loosening that results from the different annular cavities in the passage of the shield, i.e.
over-cut, due to the tools at the periphery of the cutting
wheel (the statement applies to incoherent earth).
To summarize, the pressure in the working chamber of the TBM must be maintained at a level capable
of ensuring stable working conditions with a safety
coefficient that allows the absorption of the fluctuations resulting from a dynamic situation. This should
be sufficient to avoid uncontrolled collapses of material within the working chamber but not so that it
causes soil deformations and blow-up with a consequent loss of the stabilization conditioned mixture far
from the face (AFTES 2001). The definition a priori
of the face stabilization pressure would, in fact, be a
secondary element only guided in choosing the initial value; it is optimized during the process of work
on the basis of the balance of the excavated soil volume. The problems relating to the subsidence require
maximum attention also to other parameters, in particular advance rate and backfilling operations behind
the face. Research, more technological than theoretical, must also analyse the mechanisms that rise ahead
of the face: grouting systems above the shield have
been recently introduced, mainly using polymers or
bentonite muds. These muds, injected at pressure, hold
the soil above the shield for the time period from
the passage of the head to the unthreading of the
shield and the related final backfilling with cement
mortar.
367
With regard to the five points listed above:

the component relating to the stress relief of the soil
at the face with the use of pressurized face tunneling
machines becomes unimportant if the pressure at
the face is adequately controlled.
the volume loss during the passage of the shield
is difficult to counteract and can be the weakest
element; the limitation of this component is above
all to reduce the technological over-cutting to the
minimum possible. In recent years, shield machines
that allow annular grouting for filling the region
around the shield have been used. It can also be
added that holding an high advance rate brings a
significant reduction in this component, preventing
that the gap closes completely.
the volume loss around the lined section of tunnel is minimized by using cement mortar from tail
of the shield; however problems may still occur in
the execution which render this component critical (insufficient grouting pressures, washing away
of the mixture etc.). In soil that has a self-stable
capacity adequate to the advance rate, the injection of cement mortar can create a recompression
of the ground and allows the recovery of a part of
the volume defined in point 2.
is in general unimportant compared to others.
is significant in soil subject to consolidation.
Figure 4. EPB Shield _ Individual factors of volume loss

VL.
5 VOLUME LOSS VL
The volume loss VL represents the fundamental
index parameter for the valuation of the surface settlement and, consequently, of the disturbance caused
by the tunnelling. The total volume loss is derived
from the sum of various components whose relative
contribution depends on the mechanical and physical
characteristics of the soil, on the technical characteristics of the tunnelling equipment, on the execution
method, with particular attention to the pressures
applied at the face and around, and on the advance rate.
These factors are essentially (figure 4) (Mair 1997):
volume loss at the face (face loss): owing to the
deformation of the portion of soil ahead of the face
of the tunnel and to the loss resulting from localized micro-stability problems. This loss of volume
is counteracted by the face stabilization pressure.
volume loss around the shield essentially due to
the convergence of the tunnel profile towards the
extrados of the shield machine. This factor can
be subdivided into the following terms: (2a) overcutting due to greater diameter of the tunnelling
cutterhead compared to the shield machine, necessary to reduce the friction of the shield machine
on the soil and to facilitate steering; (2b) planoaltimetric characteristics of the trail; the presence of
narrow curve ranges results in over-cutting values
higher than those present in the straight sections.
This aspect is accentuated by possible misalignments caused by the operator who may let the
machine zigzag; (2c) over-cutting caused by possible cone shaped of the shield; (2d) roughness of the
shield which can lead to movements and subsidence
due to the friction against the soil.
volume loss around the lined tunnel section due to
the gap between the tunnel profile and the extrados
of the pre-cast segments; this results in a tendency
to converge towards the extrados of the lining.
volume loss due to deformations of the lining.
long-term effects/consolidation.
Mair showed that, in the case of Slurry Shields or

EPB-S, possible references are VL 0.5% for sandy
soil and VL = 1 2% in soft clay (Mair 1997). The
definition of a reliable value is, however, very difficult in design phases because this parameter depends
on factors that are highly dependent on the specific
context and cannot be generalized (soil, technological
factors of tunnelling equipment, workmanship skill,
hydrogeology etc).
EPB-S TUNNELLING IN MILAN
Between 2004 and 2005, parallel to the construction

of the new trade fair in Rho/Milan, line 1 of the subway was extended. The track, with a length of around
2.1 km, is characterized by the presence of high water
table conditions (up to 12 m) and variable tunnel depth
between 10 and 20 m; two tunnels with one single platform were excavated using an EPB-S machine. This
method of tunnelling was previously used in Milan in
1993/1994 during the construction of the Passante
Ferroviario (underground railway line) for a section
of around 4 km when digging two tunnels with variable
depths between 4 and 16 m.
Some details of the two works are reported by
Chiorboli and Marcheselli (Chiorboli, Marcheselli
1996) and by Cavagna and Chiorboli (Cavagna,
368
Chiorboli 2004); the next table reports the main

data:
LINE 1
PASSANTE
FERROVIARIO
CLIENT
MUNICIPALITY MUNICIPALITY
OF MILAN
OF MILAN
REGION OF
LOMBARDY
BUILDING
TORNO ICLET TORNO CMB
CONTRACTORS FIAT
COGEFAR
ENGINEERING IMPRESIT
LODIGIANI
COLLINI P&C
TETTAMANTI
PERIOD
20042005
19941995
Figure 5. Line 1: settlements face pressure.
The Milan area is characterized by fluvio-glacial

deposits consisting of a heavy deposit (50 60 m) of
gravel and sand with a medium to high density. The
typical, grain size distribution, which can be defined
as the average granulometric curve is:
D < 0.074 mm = 10 20%
D > 2 mm = 70%
D > 10 mm = 10 20%
7
FACE STABILIZATION PRESSURE
The maximum settlement values are always shown in

association with the lowest possible pressure values, in
terms of minimum pressure value peaks. These minimum pressure peaks, if just momentary, reach values
significantly lower that the values required for the stability and they are, almost always, appeared where the
conditions around (exit or entry of the machine in
the construction pit,) or operating conditions (initial
start phase, mechanical rests) cause a loss of pressure that is difficult to control. If you consider this
in terms of average recorded pressure value, you cannot define an effective link between the increase in
pressure and the reduction in final settlement. The
two following figures show a summary of the average recorded pressure values, the theoretical pressure
values in K0 conditions and the measured settlement
for the two cited works. You can see that by assuming
a pressure value close to the theoretical value in K0
conditions as the start value of the machine, the pressure is subsequently reduced: the measured settlement
does not show relative associated increases.
8 TRANSVERSE AND LONGITUDINAL
TROUGHS
The transverse and longitudinal settlement troughs
have been analysed in both cases. The troughs in areas
Figure 6. Passante Ferroviario: settlements face pressure.
disturbed by the presence of buildings or grouting

treatments have been ignored. With regard to the case
of line 1 and the underground railway line, n. 18 and
n. 5 transverse settlement troughs together with n. 78
and n. 28 longitudinal settlement troughs have been
analysed. The graphic in figure 7 plots the values of
the position of the transverse point of inflection i
relating to the tunnel depth and the corresponding
interpolating lines; the same graphic also shows the
values compared with the range of literature indicated
by Mair for sand and gravel (Mair 1997). Note that for
both works considered, there is a good concentration
of data around an identifiable value such as:
It is possible to conclude that for the soil of the

Milan area the reference range K for the calculation of
settlements with the Gaussian formula is:
Figure 8 shows, for both cases, the longitudinal subsidence profiles with settlements normalised to the
369
apart from rare cases, was similar for all points

irrespective of the coverage and other parameters.
CONCLUSIONS
The analysis of the data collected and of the quoted

references enables to stress that:
1 for the Milan area soil the reference range of
the parameter K for settlement calculation is
0.43 0.46.
2 The net difference in behaviours between the two
cases essentially refers to technological aspects of
backfilling operations for the gap above the pre-cast
segments. In line 1 a simple system of grouting from 4 nozzles placed around the shield was
used. In Passante a more complex system was
adopted (Hochtief system) which uses injections
of a cement grouting from nozzles placed around
the shield; however, this bentonite fluid was able to
exert a pressure on the soil, because it was confined
to a mobile annular mould with transducers. This
equipment enables the variation of pressure of the
grouting and works in a similar way to compensation grouting, however, it is extremely taxing in
terms of maintenance and the requirement of highly
specialized personnel.
3 The analysis of the development of the longitudinal
subsidence curve over time in relation to the position of the face and the advance rate showed that,
in the Milan soil, the strain effects of the tunnel
advance spread very quickly, almost immediately,
over the surface.
4 Given the values necessary to guarantee the face
stability, the tunnel face pressure would not influence the final settlements The pressure increase
leads probably to higher safety factors of the face
stability but it is not able to guarantee comparable settlements decrease. Moreover, as explained
by Anagnostou and Kovari (Anagnostou, Kovari
1996), if the support pressure exceeds a certain
upper limit some operational problems (high wear
of the cutter, excessive torque increase, difficulties in muck discharge) may occur; it may also
cause high fluctuations of the distribution of the
effective pressure acted on the face that may lead
also to local instability. So support pressure should
be defined using only calibration during the work
starting from an initial value determined using simple equilibrium ratios in K0 conditions (AFTES
2001).
Figure 7. Transverse point of inflection tunnel depth.
Figure 8. Normalized longitudinal settlement troughs.
maximum settlement (i.e. ratio of settlements to maximum settlement is presented) the abscissa shows the
position of the face, normalized to the diameter. In
terms of absolute settlements, they show similar values in the two cases while the two percentage curves
show different values.
For the line 1 it is observed that around 40% of
the subsidence is due to the passage of the shield (9 m
at the back of the head) and, considering that around
25% is ahead of the face, the remaining 35% is behind
the machine in grouting phases for the annular space
in the extrados of the segments. This would appear to
show good management of the pressure in the tunnel
working chamber although the subsidence percentage
behind the shield appears rather high. For the Passante Ferroviario, you can see that at the tunnel face
an average value of 70%, of the subsidence, is reached;
the percentage becomes 94% with the passage of the
shield. After the passage of the face, the behaviour,
REFERENCES
A.F.T.E.S., 2001. Synthese Eupalinos 2000 EPB Shield
Theme B1: Laboratory studies on reduce models.
370
Anagnostou G., Kovari K., 1996. Face stability conditions with earth-pressure-balanced shields.Tunnelling and
Underground Space Technology, 11(2):165173. Elsevier
Science.
Attewell P. B., 1978. Ground movements caused by tunnelling in soil. Large Ground Movements and Structures, 1978 London, 812948. Ed. Geddes, Pentech Press
London.
Burland J. B., Wroth C. P., 1974. Settlement of buildings and
associated damage. Proceedings of a Conference on Settlement of Structures, 1974 Cambridge, 611654. Pentech
Press London.
Burland J. B., 1997. Assessment of risk damage to buildings
due to tunnelling and excavation. Earthquake Geotechnical Engineering, 1997 Ishihara, 11891201. Balkema.
Chambon P., Corte J. F., 1994. Shallow tunnels in cohesionless soil: Stability of tunnel face. Journal of Geotechnical
Engineering, 120, 11481165. ASCE.
Chiorboli M., Marcheselli P., 1996. Analysis and control of
subsidence due to Earth Pressure Shield tunnelling in Pas-
sante Ferroviario of Milano. North American Tunneling,

1996 Washington DC, 97106. Balkema.
Cavagna B., Chiorboli M., 2004. The use of EPB in the Milan
subsoil: line 1 extension. International Congress on mechanized tunnelling. 2004 Torino, 107118. Politecnico
Torino.
Mair R. J., Taylor R. N., Burland J. B., 1996. Prediction of
ground movements and assessment of risk of building
damage due to bored tunnelling. International Symposium
on Geotechnical Aspects of Underground Construction in
Soft Ground, 1996 London, 713718. Balkema.
Mair R. J., Taylor R. N., 1997. Theme lecture: bored tunnelling in the urban environment. Conference on soil
mechanics and foundation engineering, 1997 Hamburg,
vol. 4, 23532385. Balkema.
Rankin W. J., 1988. Ground movements resulting from urban
tunnelling: predictions and effects. Engineering geology
of underground movements, Geological Society Engineering Special Publication, 7992. Geological Society,
London.
371
Displacements and stresses induced by a tunnel excavation: Case of

Bois de Peu (France)
S. Eclaircy-Caudron, D. Dias & R. Kastner
INSA Lyon, Civil and Environmental Engineering Laboratory (LGCIE), Villeurbanne cdex, France
ABSTRACT: The tunnel of Bois de Peu is part of a project of the south-eastern Besanon (France) by-pass.
An exploration gallery permitted to highlight the presence of eighteen geological units and to distinguish four
sorts of materials: limestone, marls, clays and interbedings of marls and limestone. Despite of the important
number of laboratory and in situ tests carried out, many uncertainties remained on the mechanical parameters
value and on the position of the different geological units. So, it was decided to apply the interactive design method
during the construction to adapt the excavation and support to the actual conditions found. In the framework
of this method, an important monitoring program was foreseen. This article shows the behavior observed in the
different kind of soils. In the clayey zone where the support is the most complex, strains in the ground and in steel
ribs are monitored. Finally, the available measurements permit to better understand the behavior of soil-structure
interaction problems.
INTRODUCTION
The digging of a tunnel induces a modification of the

initial stress field in the ground which creates an unbalance state. This unbalance results in movements of soil
like convergence of the cavity, pre-convergence ahead
of the face, extrusion of the face and settlements.
During the excavation an arching effect is created. If
it is not sufficient to stabilize the cavity, lining supports
are set up in order to limit soil movements and thus, to
avoid the failure of the structure.
The lining support can include shotcrete, ribs and/or
radial bolting. In certain cases, when the ground has
low mechanical characteristics, additional support like
ground reinforcement or ground improvement at some
stage of the excavation is necessary. Several reinforcement systems have been developed. Pelizza & Peila
(1999) presented the different methods of soil and rock
improvement used to permit safe tunneling in difficult geological conditions. Lunardi (2000) divided the
support methods into three groups: pre-confinement,
confinement and pre-support. Each group exerts a
different kind of effect on the cavity.
When the traditional method of digging is used in
difficult geological conditions, the main problem is
the control of movements. Without support or adapted
treatment, the ground tends to sink into the opening
(tunnel face failure, tunnel face extrusion): it is the
phenomenon of decompression. In order to reduce this
phenomenon, an action of pre-confinement may be
required. A pre-confinement action is defined as any

active action that increases the formation of an arch
effect in the ground ahead of the tunnel face. The
pre-confinement can be achieved by reinforcement
or protective intervention ahead the tunnel face. The
umbrella arch method and the face bolting are included
in protective interventions or pre-support methods.
The support and forepoling introduce many three
dimensional soilstructure interactions. Consequently,
it is difficult to understand these phenomena analytically. Moreover, although the umbrella arch method
is widely used, there are no simple approximations
to simulate this method in numerical analyses. The
design of an umbrella arch is still based today on
empirical considerations or on simplified schemes
(Oreste & Peila 1997). Several numerical studies carried out in 2D and in 3D, were focused on the manner
of taking into account umbrella arch (Tan & Ranjith
2003, Bae et al. 2005). In the same way, several numerical modeling were carried out in 2D and in 3D on the
manner of taking into account the face bolting (Yoo
2002, Dias 1999).
For some large geotechnical engineering projects,
a monitoring program is generally defined in project
phase to record the soil movements which really occur
during construction and to evaluate the performance
of the construction design. In most of the cases, the
recorded data are just used to control the construction
process. But, these data can be also used to update
predictions by using inverse analysis processes on
373
Figure 1. A plan view of the Bois de Peu tunnel project.
these measurements in order to decide of a possible adaptation of the construction process in the case
where unsafe values would be predicted. This later
practice is a part of the observational method (Peck
1969, Powderham & Nicholson 1996). The AFTES
guidelines (2005) related to the monitoring methods
of underground works present the major methods and
give advices on the measurements frequency.
In this article, monitoring results obtained in the
different geological units found during the excavation
of the downward tube of the Bois de Peu tunnel are presented. First, this communication introduces the tunnel
project. Then, in a second part, the monitoring results
are presented. They permit to better understand the soil
and the umbrella arch behavior set up in the clayey
zone.
2
2.1
PRESENTATION OF THE TUNNEL

General presentation
The tunnel of Bois de Peu (cf. Fig. 1) is part of

the project of the south-eastern Besanon (France)
by-pass entitled La Voie des Mercureaux. This
project includes several engineering structures: two
tunnels and one bridge, and several retaining walls.
The tunnel is composed of two tubes of 520 m length.
The cover height varies between 8 m and 140 m. The
excavation, achieved in September 2006, was carried
out full face by drill and blast for the major part of the
tunnel.
2.2
Geology and geotechnics
An exploration gallery was dug in 1995 in order to

assess the mechanical properties of the ground. It has
a width of 3 m and a height of 3.5 m. Various laboratory and in situ tests were carried out. The results lead
to conclude that the tunnel is situated in a disturbed
area. Eighteen geological units are identified. A geological cross section is showed in Figure 2. Among
these eighteen units, four sorts of materials can be
distinguished: clays, marls, limestone and interbedings of marls and limestone. Geotechnical properties
of these materials defined at the end of site investigations are summarized in Table 1. Two types of
characteristics were defined for the marl (probable and
Figure 2. The geological cross section.

Table 1.
Geotechnical properties of material.
Material
Limestone
Marl
(probable)
Marl
(exceptional)
Clay
kN/m3
E
MPa
MPa
26.3
24.8
329010000
1600
0.7
40
24.0
750
0.21
36
23
80
0.250.4
1317
exceptional) because the in situ and laboratory tests

leaded to variable values. The Poisson ratio, the dilatancy angle and the earth pressure ratio are the same
for the materials and are respectively equal to 0.3, 0
and 0.7.
The exact position of the different units is difficult to
know before the digging. Finally, following the observation made in this exploration gallery, many uncertainties are remaining. So, it was decided to apply the
observational method during the digging in order to
adapt the lining support to the real ground conditions.
In this framework, an important number of experimental measurements were foreseen and four sorts of
support were defined in project phase. During construction, the choice of the adapted support depended
on the monitoring results and on the quality of the
soils assessed by geological surveys. In this context,
two other supports were realized during the excavation. The digging step for each support was defined
variable.
3
MONITORING RESULTS
Four monitored sections are presented: one in interbedings of marl and limestone (entitled D1), one in
marls (D2) and two in clay (D3 & D4). Each of them
are not circular. For two of them, only convergence and
leveling measurements are available. For the others,
more specific measurements were carried out such as
strains measurements ahead the face by extrusometers.
374
Table 2.
Main characteristics of each studied section.
Section
PM
m
Material
H
m
Dface
m
D1
D2
D3
D4
9.75
457
510.3
493.8
marl/limestone
marl
clay
clay
22
40
15
22
3.1
13
1.3
3.84
Each section is presented and analyzed. Table 2 reports

the localization in the tube (PM, referred to Fig. 1),
the type of material, the overburden from the tunnel
axis (H) and the distance from the face at the origin of
the convergence and leveling measurements (Dface ) for
each studied section. D1, D3 and D4 are situated near
one of the two tunnel portals (see Fig. 1, low overburden). For the sections D1, the wall support is composed
of shotcrete and steel ribs set up every 0.75 to 1.75 m.
For D2, it is made up by shotcrete and radial bolts. The
diging step varies between 3.5 and 4.5 m. For D3 and
D4, the support is more complex. It includes a wall
support and an arch invert by shotcrete and steel ribs,
set up at the tunnel advance every 1.5 m. And a forepoling by umbrella arch and face bolting are realized
every 9 m in the general case. In D3, the excavation
was made in partial face otherwise it was made in full
face in D4.
3.1
Figure 3. Leveling measurements in the half higher

section D3.
Figure 4. Convergences measurements in the half lower

section D3.
Convergence and leveling
Tunnel wall convergences between reference points

are realized by optical sights. Five optical reflector targets are installed in the monitored sections where the
excavation is made in full face (at the crown, at 45
and at the spring line). When the excavation is realized
in partial face, the five targets set up after the excavation of the half higher section are completed by two
other targets installed after the excavation of the lower
part at the side walls. Leveling measurements are less
accurate than convergences (5 mm against 1 mm)
because of the use of several reference stations. But,
for all studied sections in the downward tube, leveling
curves are exploitable.
The maximal convergence and leveling values are
obtained in the clayey zone. In the half higher section
D3, leveling measurements reached 25 mm for targets
1 and 2 (Fig. 3) while convergences remain lower than
7 mm.
The maximal convergence value is recorded for
wire 4. Although the section is closed by a temporary
buton, vertical movements are important. As targets 1,
2 and 3 present higher displacements, the deformation
of the section is dissymmetric. The side wall located
near the other tube presents larger vertical movements.
After the excavation of the lower part, convergences
measurements are higher and reach 12 mm for wire 4
(Fig. 4). For the other wires, convergences remain

lower than 7 mm. The leveling measurements are not
so high than in the half higher section. They remain
lower than 6 mm. All targets show similar displacements so the deformation of the section is symmetric
after the excavation of the lower part. Measurements
in section D3 seem to be stabilized after the return
towards an excavation in full face.
In section D4, 25 mm of convergence is registered
for wire 4 (Fig. 5) and the speed of convergence is
high despite of the distance from the face at the origin which is important (4 m). Vertical movements are
important but remain lower than those registered during the digging of the half higher section D3. They
reach 14 mm. The section is closed by an arch invert
set up at the tunnel advance every 3 m. So, the section
is not immediately closed. This can explain the leveling values. Measurements are not stabilized at the end
of the excavation.
Displacements recorded in the clayey zone in the
downward tube are more important than those measured in the other tube (Eclaircy-Caudron et al. 2007).
Consequently, the clayey zone seems to be of better
quality in the downward tube. This conclusion is also
375
Figure 5. Convergences measurements in section D4.

Figure 7. Face survey in section D1.
Figure 6. Leveling measurements in section D1.

Figure 8. Face survey in section D2.
verified by the strains measurements performed by

extrusometers ahead the face.
The section D1 is situated in interbedings of marls
and limestone. The convergence measurements show
an important dissymmetry confirmed by the leveling
measurements (Fig. 6). In fact, vertical displacements
are more important for targets 4 and 5 and reach 4 mm.
Maximal convergences values are obtained for wires
3 and 4 and reach 8 mm.
This section presents lower measured values than
sections situated in a similar soil in the upper tube
(Eclaircy-Caudron et al. 2007). This can be explained
by the dip of the interbedings which is more favorable in the downward tube (Fig. 7). In fact, the face
survey shows almost horizontal interbedings of marls
and limestone. Moreover, the face survey presents
less fractures and faults in this section than in similar
sections in the other tube. So, the soil seems to be of
better quality in the downward tube. Measurements are
stabilized at a distance from the face equal to 60 m so
9R against 150 m in the other tube.
In section D2, located in marls, the maximal convergence value is obtained for wire 4 as in the section
D1 and in marls in the upward tube. It reaches 12 mm
before the end of the excavation from the Doubs portal
at 03/17/06. Leveling measurements show a dissymmetric deformation of the section as in the upward
tube. Maximal vertical movements are recorded for

targets 3 and 2 and reach 15 mm before the end of
the excavation from the Doubs portal and almost
20 mm after. The side wall situated on the side of
the other tube presents more displacements than the
other while in section D1 the opposite is noticed. This
dissymmetry can be explained by the presence of the
exploration gallery and not by the geology which is
perfectly symmetric (Fig. 8). Convergences measurements are stabilized before the end of the excavation
from the Doubs portal (at 25 m). After the excavation
from the Doubs portal and before the beginning of the
excavation from the Vallon portal, the measured convergences increase. These displacements increments
can be due to the construction phases (drilling of the
first face bolts and of the first umbrella arch). After
the beginning of the excavation from the Vallon portal,
the convergence measurements stop increasing. So, the
digging from the other portal does not influence the
convergences measurements of this section while in
the other tube the opposite was noticed in this kind
of soil. Concerning the leveling, no conclusion can be
drawn because after the beginning of the excavation
from the Vallon portal, these measurements present
some unexplained variations.
376
Figure 9. Extrusion evolution versus the distance from the

face at different face positions (extrusometer 1).
Figure 10. Extrusion evolution versus the PM at different

face positions (extrusometer 1).
3.2
The length of the extrusometer is indicated in each

case. Four measurements were performed when the
face was stopped at PM 510 after collapses which took
place in May 2006. After the resumption of the excavation in partial face only one measurement was realized
and it is not exploitable due to the short length of the
extrusometer. The maximal value of extrusion reaches
10 cm (so a strain U/R equal to 1.5% with R = 6.8 m).
This value is four times greater than in the other tube.
80% of this extrusion is obtained in the first 4 m so at a
distance lower than one radius as in the other tube. Due
to a lack of measurements, it is not possible to determine the radius of influence of the face. Measurements
should be realized at each digging phases.
Radial displacements in marl at PM 325
These measurements are realized by three borehole

extensometers installed radially from the tunnel wall
(at the crown and at 45 ). Each extensometer has a
length of 12 m and included 7 measurements points at
0, 1, 2, 4, 6, 8 and 10 m from the tunnel wall. These
measurements can be used to assess the extent of the
zone of influence around a tunnel. Movements are
measured automatically with an accuracy of 0.02 m.
The face was at a distance of 4 m at the origin of the
measurements. The point located at 10 m is assumed
to be outside of the influence zone. So, the displacement of each point is computed by considering that
this point is fixed. The computed displacements can
be compared to the convergences measurements realized in marls. Displacements are very low compared
to convergences and remain lower than 2 mm. The distance from the face being equal to 4 m at the origin of
displacements measurements, low displacements can
be explained by the fact that a part of them was lost.
3.3
Displacements measured ahead the face in the

clayey zone
Two extrusometers of 20 meters length were set up

at PM 521 and 501. The feature consists to measure relative displacements between two successive
points spaced by one meter. It is destroyed at the
tunnel advance. If the anchor point can be considered outside the zone influenced by the excavation
then absolute displacement of each point can be computed. Generally, the zone of influence extends until
one radius ahead the face. So, to consider that the
last point is fixed, the extrusometer has to be longer
than one radius. The measurements obtained for each
extrusometer are presented.
3.3.1 Extrusometer 1 (PM 521)
Figures 9 and 10 present the extrusion evolution versus
the distance to the face and the PM.
3.3.2 Extrusometer 2 (PM 501)

Figure 11 shows the most important extrusion measurements performed with the second extrusometer.
Fourteen measurements were recorded. One was made
during the excavation of the half higher section. Four
were performed between the end of the digging of the
half higher section and the beginning of the excavation
of the lower part. Four were realized after the beginning of the lower part excavation and three were made
after the complete excavation of the part in partial face.
Finally, only two measurements were realized after the
return towards an excavation in full face. The curves
evolution versus the distance from the face is different.
This can be explained by a different behavior of soil or
of the bolts. The maximal extrusion reaches 4 cm (so
U/R = 0.7%). This value is similar to the one obtained
in the other tube.
3.4 Strains in the steel rib at PM 493

These measurements are realized by 14 vibrating wires
extensometers installed on steel ribs as shown in
Figure 12. They permit to obtain the strain of the
steel rib. Movements are measured with an accuracy
of 1 m/m. Temperatures are also monitored. The face
377
Figure 11. Extrusion evolution versus the distance from the

face at different face positions (extrusometer 2).
Figure 13. Evolution of the stresses in the steel rib at

PM 493.
Figure 12. Location of the instrumentation at PM 495.
was at a distance of 1 m at the origin of the measurements. Maximal values are recorded in extensometers
S2-CV1 and S2-CV2 and reach 400 m/m. So, the
deformation of the rib is dissymetric. However, the
geology is the same in both sides. From the strains
measured by a pair of extensometers it is possible to
compute the maximum stress induced in the steel rib.
For example, the stress S12 is determined from the
measurements of S2-CV1 and S2-CV2. These stresses
are important and the maximal value is obtained
in extensometers 1 and 2 and reaches 100 MPa.
However, this value remained lower than the limit
stress (180 MPa). The computed stresses are shown in
Figure 13. The same behavior than in the other tube
is observed. Stresses induced in the steel rib permit
also to compute the normal force and the bending
moment. It appears that the steel rib works essentially
in compression as in the upward tube. As regard to
the measurements, structural elements are less loaded
in the downward tube than in the other tube but
displacements are higher.
CONCLUSIONS
This paper presents the most representative measurements which occurred during the excavation of the
downward tube of the Bois de Peu tunnel in Besanon
(France). The measurements permit to better understand the behavior of the complex support set up in
the clayey zone and to adapt the support and excavation to the found conditions during the construction.
The convergences measurements vary between 6 and
25 mm and the leveling between 4 and 25 mm following the soil. It appeared that structural elements are
less loaded in the clayey zone of the downward tube
than in the upward tube but displacements are higher.
In fact, vibrating wires extensometers set up on the
steel rib permit to conclude that this structural element is loaded at 55% of its admissible stress against
70% in the upward tube where the clayey zone seems
to be of better quality.
In the upward tube, higher displacements are
observed in the geological unit including interbedings of marl and limestone where a heterogeneous
geology is observed with many faults. In the downward tube, this unit is less fractured and displacements
are reduced. The displacements monitored by extrusometers show that the perturbed zone ahead the face
extends to one diameter. And, 80% of the extrusion at
the face is obtained at a distance lower than one radius.
Comparisons between predictions and measurements
permitted to evaluate the actual mechanical parameters of the soils found by the tunnel. Comparisons
in the clayey zone showed that this layer had similar properties than those defined at the end of site
investigations.
REFERENCES
AFTES. 2005. Recommandations relatives aux mthodes
dauscultation des ouvrages souterrains. Tunnels et
Ouvrages Souterrains 187: 1047.
378
Bae, G. J., Shin, H. S., Sicilia, C., Choi, Y. G. & Lim, J. J.

2005. Homogenization framework for three dimensional
elastoplastic finite element analysis of a grouted piperoofing reinforcement method for tunnelling. International Journal for Numerical and Analytical Methods in
Geomechanics 29: 124.
CETU. 2003. Dossier de consultation aux entreprises. Tunnel
de Bois de Peu.
Dias, D. 1999. Renforcement du front de taille des tunnels par
boulonnage Etude numrique et application un cas rel
en site urbain. Thesis, Institut des Sciences Appliques de
Lyon, Lyon, France.
Eclaircy-Caudron, S., Dias, D. & Kastner, R. 2007. Movements induced by the excavation of the Bois de Peu tunnel
in Besancon (France). ECCOMAS thematic Conference on
Computational Methods in Tunneling (EURO:TUN 2007
Vienna, Austria, 2729 August 2007).
Lunardi, P. 2000. The design and construction of tunnels
using the approach based on the analysis of controlled
deformation in rocks and soils. Tunnels and Tunnelling
International: 330.
Oreste, P. P. & Peila, D. 1997. La progettazione degli infilaggi

in avanzamento nella costruzione delle gallerie. Convegno
di Ingegneria Geotecnica; Proc. Intern. Symp., Perugia.
Peck, R. B. 1969. Advantages and limitations of the Observational Method in applied soil mechanics. Gotechnique
19: 171187.
Pelizza, S. & Peila, D. 1999. Soil and rock reinforcement in
tunneling.Tunnelling and Underground SpaceTechnology
8: 357372.
Powderham, A. J. & Nicholson, D. P. 1996. The Observational
Method in Geotechnical Engineering. Institution of Civil
Engineers: 195204. London: Thomas Telford.
Tan, W. L. & Ranjith, P. G. 2003. Numerical Analysis of
Pipe Roof Reinforcement in Soft Ground Tunnelling.
16th International Conference on Engineering Mechanics, ASCE, Seattle, USA.
Yoo, C. 2002. Finite element analysis of tunnel face reinforced
by longitudinal pipes. Computers and Geotechnics 29:
7394.
379
Shield tunneling beneath existing railway line in soft ground

Q.M. Gong & S.H. Zhou
The Key Laboratory of Road and Traffic Engineering at Tongji University, Shanghai, P.R. China
ABSTRACT: The paper discusses the geodynamic challenges and technical countermeasures in the shield
tunnel projects passing under an existing railway line, especially for soft ground conditions. Three main issues
needed to be solved by the engineering community: train safety during the period of tunnel construction, accelerated track degradation of track structure in the long term and tunnel structure durability. This paper describes
and predicts the nature of the three issues, the countermeasures are also given.
INTRODUCTION
Table 1.
Cities and urban population steadily grow, and put

stringent requirements for metro transportation. With
the development of the metro system, more and more
metro tunnels need to pass under an existing railway line, such as metro line 9, 11 and 7 in Shanghai,
China. At the same time, the railway lines have been
speeded up and are planned for increasingly higher
train speeds and higher loads, while existing lines need
to be upgraded to allow faster, heavier and more frequent trains. In this context, three main issues need to
be considered:
1 The Train Safety issue: excessive settlement of
the track due to the tunnel excavation must be
avoided to ensure the permissible deviations in
track geometry.
2 The Track Structure Durability issue: increasing traffic density, loads and metro tunnel structure bring about higher dynamic actions and more
intensive track maintenance.
3 The Tunnel Structure Durability issue: dynamic
actions are to be sustained by tunnel structure.
This paper discusses these issues in Shanghai,
China. The train safety issue and countermeasures
during the metro tunnel construction are concentrated.
2
SITE DESCRIPTION
The site chosen for the project was the Jiading area in
Shanghai, where metro line 11 passes under Hu-Ning
railway Line. Figure 1 shows their plane and vertical
position. The angle between them is 85 degree, almost
perpendicular. The overburden soil depth of shield tunnel is 11.08 m, its outside diameter is 6.2 m and the
thickness of tunnel lining is 0.35 m, the groundwater
level is about 1m below the ground surface.
Soil parameters.
Soil layer
Parameters
Gravity ()
(kN/m3 )
Cohesion C(kPa)
Angle of friction
(degrees)
Lateral ratio of
earth pressure K0
Compressive
Modulus (MPa)
18.4
18.1
17.4
17.7
19.3
13.0
18.5
8.0
24.0
9.0
17.0
13.0
13.0
45
15
0.46
0.41
0.50
0.50
5.25
7.22
5.50
6.77
0.40
10.5
* In Table 1, 1 is Tan silty clay, 2 is Sallow silty clay, 1 is

grey silty clay, is grey clay, is Sap green Clay.
The 4.5 m high embankment of Hu-Ning railway

line has been constructed in 1908. The train speed
have increased up to 250 km/h at 18th, April, 2007
and demands for shortening travel times are rising.
Figure 1(b) also shows the geological profile of the
chosen site, the soil properties are in table 1. According
to the geotechnical investigation, the site can be characterized by a 1.5 m weathered crust over a layer of soft
clay with a thickness of about 13.0 m. Under these layers lies silty clay whose stiffness is more greater than
the overlying soils.
Track-side buildings are almost not existing, so the
influences on environment can be ignored.
3 THE SAFETY OF RUNNING OF THE TRAINS
DURING TUNNEL EXCAVATION
Tunnel excavation may induce adverse effects on
nearby existing structures and services (e.g., deformation on tracks, derail of trains). The accurate prediction
381
(a) Plain position
(b) Geological profile
Figure 1. Layout of railway line and shield tunnel.
of tunneling effects poses a major challenge during

design.
The effects of tunnel advancement during construction on the ground and railway track are three
dimensional and transient. In this paper, a 3D, elastoplastic analysis was performed. Figure 2 shows the 3D
finite element mesh. The mesh was 50 m long, 45 m
wide and 40 m high. Eight-node brick elements and
four-node shell elements were used to model the soil
and the concrete lining, respectively. Roller supports
were applied on all vertical sides of the mesh, whereas
fixed supports were assigned to the base of the mesh.
Therefore, the movement normal to all vertical sides
of the mesh and the movement in all directions at
the base of the mesh were restrained. The water table
was located at the ground surface. An elasto-perfectlyplastic soil model using the Mohr-Coulumb failure
criterion with a non-associated flow rule was adopted
in this study. The tunnel lining was modeled as a linear
elastic material. The Youngs modulus and Poissons
ratio for the tunnel lining were taken as 30 GPa and
0.3, respectively. The unit weight of the tunnel lining
was 24 kN/m3 .
Figure 3 shows the progressive changes in the
tunneling-induced surface settlement under the track
as the tunnel advances. With further excavation, the
surface settlements continue to increase. Considering
the tunneling-induced deformation, the irregularity of
the track is shown in Figure 4.
Figure 2. Finite element mesh.
The derail coefficient and rate of wheel load reduction were shown as table 2 using locomotive-track
dynamic coupling model. According to the protocol
published by the Ministry of Railway in China, the
stability and safety of the train will be threatened if
the running speed of train is above 100km/h (passenger car) and 60 km/h (freight), so the train speed must
be limited during the tunnel construction.
3.1 Influence on tunnel lining by running trains
After the construction of the tunnel, the dynamic
stresses induced by a running train will act on the tunnel lining for a long time, even until the railway line
or the metro line will be abandoned. So the influence
of the dynamic stresses must be considered. In general, the dynamic stresses induced by a train running
dissipate quickly because of the nonlinear material
and the viscous damping. The influence depth is only
about 3m under the sub-grade bed. But the values of
dynamic stresses will be greater with the tunnel under
382
p13
Surface settlement (mm)
5
m
p11
p12
0
0
-5
10
15
20
25
30
35
40
45
P1
50
q11
q1
q12
q1
-10
-15
O
-20
-25
q21
Uplink tunnel excavates 10.6m

q2
q2
q22
P2
p12
p11
p13
Figure 5. Load-structure model including dynamic stress.
dynamic stresses. It is quite different from the conditions with only static stresses. The long-term actions
of dynamic stresses will influence the service life time
of tunnel structure.
Figure 3. Tunneling-induced surface deformation.
SAFETY COUNTERMEASURES
4.1 Countermeasures to ensure the train safety

during tunnel excavation
Figure 4. Longitudinal irregularity of the tracks by
tunneling-induced deformation.
Table 2. Maximum value of derail coefficient and wheel
load reduction.
Vehicle
Passenger car
Freight
Speed
(km/h)
90
50
Item
Pair of I
wheels II
III
IV
0.63
0.51
0.48
0.60
0.13
0.10
0.08
0.11
0.75
0.59
0.56
0.73
0.20
0.13
0.11
0.21
0.69
0.56
0.53
0.67
0.21
0.15
0.15
0.19
0.98
0.78
0.75
0.93
0.25
0.22
0.20
0.23
100
60
*In Table 2, represents Derail coefficient, represents

wheel load reduction.
the railway line, because the stiffness of tunnel is much

greater then the soft soil at the same depth.
The dynamic stresses acting on tunnel lining have
been established using a dynamic numerical model.
Laboratory tests have shown that soil stiffness and
damping change with cyclic strain amplitude under
dynamic cyclic loading conditions. To simplify, the
soil is modeled using equivalent Linear analysis, the
damping ratio also changes with strain. Figure 5 shows
the loads acting on the tunnel lining including the
There are basically three ways to decrease the influence on safety of the trains during tunnel construction:
(a) by decreasing the tunneling induced deformation;
(b) by increasing the longitudinal bending stiffness
of track structure itself; (c) limit the running speed
of the train.
Decreasing the tunneling induced deformation may
for instance be done by means of improving the stiffness and strength of the soil around the tunnel. This can
be done by for example lime-cement piles, jet-piles
or other deep-mixing methods. These countermeasures are easier to implement as part of the foundation
work of new lines, than as a retrofitting method under
existing lines.
Figure 6 shows the reinforcing areas and methods
around the tunnel, including jet-piles and grouting.
These countermeasures decrease the track deformation during the tunneling. The value of the settlements
is only 15% of the original deformation, so the remaining track irregularity is small. At the same time, the
bearing capacity of the foundation of the railway track
after the tunnel construction will increase.
The construction parameters should be controlled
during the metro tunnel excavation to decrease the disturbing of the soil. Table 3 shows the recommended
parameters. The controlling and monitoring systems
should also be perfect to follow the track deformations.
Figure 7 shows the measured deformations at the site
during tunnel construction. is the preliminary settlement, is the settlement induced by shield arrival,
is the settlement during tunnel construction, is the
383
4.2 Countermeasures to protect the tunnel lining

Considering the long life time and the exhausted
strength, the lining stiffness and strength should be
increased. So the reinforcement ratio of the tunnel
lining under the railway line has been increased.
5
Tunnels passing under an existing railway line must

ensure the train safety, the track structure durability
and the tunnel structure durability. These problems
are especially important in soft ground areas. In order
to decrease the dynamic stresses acting on the tunnel
lining, to decrease the tunneling induced deformation, and to even increase the bearing capacity of the
ground, the soft soil and tunnel lining stiffness must
be reinforced.
Secondary reinforcing area
Main reinforcing area

jet-pile area
Tunnel structure
Figure 6. Reinforced soil around the tunnel.

Table 3.
CONCLUSIONS
Construction parameters.
ACKNOWLEDGEMENTS
Support pressure (MPa)

Grouting pressure (MPa)
Grouting amount (m3 )
Grouting speed (L/s)
Construction speed (cm/min)
0.20 0.22
0.19 0.21
3.3 4.1
1.1 1.4
2.0 2.5
Cumulative Settlement(mm)
Construction parameters
Construction site
Hu-Ning
Railway
This research was sponsored by National NaturalScience Foundation of China (No.50678131) and
Science and Technology Commission of shanghai
Municipality (No.06ZR14083). Grateful appreciation
is expressed for these supports.
REFERENCES
Esveld, C. 1989. Modern Railway Track. MRT-Productions,
ISBN 90-800324-1-7.
Esveld, C. 1997. Innovation for the control of infrastructure
maintenance. Rail International/Schienen der Welt.
Suiker, A.S.J. 1996. Dynamic behaviour of homogeneous
and stratified media under pulses and moving loads.
TU Delft, Report 7-96-119-1.
10
5
0
Subsequent settlement
-5
-1 0
-1 5
-2 0
0
30
60
90
120
150
180
Time (h)
Figure 7. Measured deformation at the site during tunnel

construction.
settlement induced by sub-optimal filling of shields

rear, is the subsequent settlement. The results show
that the deformation of the track has been controlled
effectively. These approaches can be used.
384
Case history on a railway tunnel in soft rock (Morocco)

A. Guiloux, H. Le Bissonnais & J. Marlinge
TERRASOL, Geotechnical consulting engineers, France
H. Thiebault, J. Ryckaert, G. Viel & F. Lanquette

SETEC TPI, France
A. Erridaoui
MAROC SETEC, Morocco
M.Q.S. Hu
TEC Engineering, P.R. China
ABSTRACT: Ras RMel Moroccan tunnel is a 2.6 km long and 60 m2 section single track railway tunnel.
It was excavated in a complex geological context of highly heterogeneous and highly deformable soft rock.
The purpose of this paper is first to describe the methodology used for specific design of support and lining.
The deformation monitoring process applied during construction will then be detailed, and an analysis of the
deformations measured will be given, making a comparison with design calculations results. Main concerns
were potential face instability and high tunnel deformations.
PROJECT DESCRIPTION
The Ras RMel tunnel presented in this case-history is

a single track railway tunnel, part of the new railway
line being constructed in Morocco between the town
of Tangier and its new Mediterranean harbour, located
on the coast about 30 km north-east of the city. The
tunnel itself is located some 15 km east of Tangier, near
the location of Ras RMel. Western portal is called
Tangier portal, eastern one is named Ras RMel
portal. Kilometric points on the railway line increase
from Tangier portal (PK 26+841) to Ras RMel portal
(PK 29+445).
Chinese company TEC Engineering won the contract for construction of the tunnel, which also included
construction of a second tunnel, shorter (600 m long)
but in a similar geotechnical context, and construction
of the railway platform in between the two tunnels.
For both tunnels, following a previous collaboration
on the Meknes tunnel project (another Moroccan railway tunnel, the study of which was described in a
previous paper for AFTES October 2005 international
congress), TEC Engineering chose French engineering
companies SETEC TPI and TERRASOL, subsidiaries
of French group SETEC, as its consulting engineers for
specific design of support and lining, and for deformation analysis during construction. This paper focuses
on design and construction issues regarding Ras RMel

tunnel.
The tunnel, of a 2600 m length and a 60 m2 section,
was excavated in highly heterogeneous flyschs, with up
to 150 m depth in its central part (Cf. Fig. 2). The excavation was given a horse-shoe shape of a 8.5 m height
and a 7.6 m to 7.8 m width, depending on the lining
thickness in order to preserve the ultimate circulation
clearance (Cf. Fig. 1). Four zones with provisional
enlarged sections were excavated, in order to allow
entering and exiting trucks to cross during construction. The enlarged sections are of a 10.2 m width and
a 9.4 m height. At final stage, the over-excavation of
those zones will be filled with lining concrete.
The entrance portals were at a depth of only 10 m.
As those portals were almost vertical, and as the quality
of the pelites, classified as soft rocks, is rather poor, it
was decided to stabilise the portals building two 22 m
long false tunnels (one at each portal) using a cut
and cover technique.
Indicative tunnel profile is given on Figure 2.
From each portal, and on a total length of around
1600 m, the excavation is at a depth lower than 50 m:
great surface settlements were expected in this zone.
In the central part of the tunnel, on a length of 1000 m,
excavation is at a depth of about 150 m: low surface settlements but high convergence and tunnel settlements
385
Table 1.
Geomechanical characteristics.
Formation
Mainly
pelite
Mainly
sandstone
Fully
sandstone
(kN/m3 )
UCS (MPa)
Cu (kPa)
u ( )
C (kPa)
( )
E0 (MPa)*
E (MPa)*
g (kPa)
24
0.66
200
0
20
28
500
250
0,3
200
24
500
0
100
30
1000
0,3
0
24
46
1000
35
1500
0,3
0
* E0 and E are respectively short and long term Youngs

moduli.
Figure 1. Type profiles (basic and enlarged cross-sections).
Furthermore, the geological formations are complex themselves: both flyschs formations consist of
alternating levels of pelites (soft rock with Unconfined
Compressive Strength (UCS < 1 MPa) and highly
resistant sandstones (UCS = 30 to 40 MPa). Four (4)
kinds of formations were then expected:
non-altered, highly fractured pelites, with occasional breccia zones, Rock Quality Designation
(RQD) ranging from 50% to 75%, and sandstones
levels RQD, expressed as a percentage, is the
summed length of core pieces greater than 10 cm
measured for a 1 m long core pass; once excavated,
the pelites get quickly altered;
blocky sandstones, with occasional pelite levels;
alternating levels of pelites and sandstones of low
thickness, that can be found intricated in blocky
sandstones;
claystone breccia.
Figure 2. Tunnel profile and deformations measured before

lining construction.
were expected in this zone. As the tunnel is in a countryside environment with no construction at surface,
surface settlements were not expected to be of a major
concern, but tunnel convergences were.
As blocky sandstones layers where not persistent, no

continuity for those levels could be drawn between the
borehole logs, and no accurate geological profile could
be built for the project. It has then been decided, to propose an indicative geological zoning along the axis of
the tunnel, for support and lining design of the project,
pointing at three main categories of formations:
GEOTECHNICAL CONTEXT
Ras RMel tunnel is located in a geologically complex

and highly tectonised zone. Two flyschs formations
can be found, the older one (the Tisiren nappe, alternating levels of sandstone and pelite) being on top of
the younger one (the Beni Ider Nappe, softer clayeycalcareous flysch) being a thrust-nappe. The contact
plane between those formations slightly dips towards
east. It was expected to be intersected by the tunnel
but could not be firmly identified during construction. This contact is known to be intersected by several
sub-vertical faults disturbing locally the lithology.
formations mainly pelitic;

formations mainly comprising sandstones;
formations fully comprising sandstones;
Table 1 sums up the average mechanical characteristics given to these equivalent formations. Short term
cohesion values for mainly pelitic formation are the
updated values following analysis of first measures of
tunnel deformations.
Potential swelling pressure g in pelites is evaluated
to 200 kPa, applying under lining invert.
Because of the thrust nappe context, high values for horizontal stresses were expected, which was
386
taken into account in the design calculations setting

parameter K0 to 1 (horizontal stress equal to vertical
one).
During construction, sandstone levels were confirmed non-persistent, which meant that the Fully
Sandstone formation was non relevant. The two other
formations describe soft rocks with rather weak characteristics: under expected geostatic stress state, the
stability ratio 0 /UCS ranges from 2 to 3,5 (>1) meaning that stress level is high enough for excavation to
be unstable without support.
From a hydrogeological point of view, captive water
of limited extent had been identified in the boreholes,
sheltered by sandstone layers. The surrounding pelite
matrix is little permeable. However, the highly fractured pelites and sandstones might allow limited water
flows to reach the tunnel. It was then decided to protect
tunnel lining with a watertight membrane on crown
and shoulders.
f = uf /R (%)
60%
55%
50%
45%
40%
35%
30%
25%
20%
15%
10%
5%
0%
1
1.2
1.4
1.6
1.8
Safety factor F
Figure 3. Tunnel face stability analysis for P1 profile

(mainly pelitic formation at 50 m depth).
TERRASOL and SETEC TPI performed specific

design studies for support and lining of the tunnel,
as well as stability analysis of tunnel face. In order to
control tunnel and surface deformations, a full face
excavation method was considered.
Based on the indicative geotechnical zoning proposed for the tunnel, six calculation profiles were
defined, named P1 to P6. Those profiles differed by
the formation and depth taken into account for calculation (each profile is supposed to be composed of
one single formation). Five additional profiles were
studied, corresponding to potential enlarged section
zones. The precise location of those zones would be
defined precisely during construction, depending on
real geological conditions.
by a security factor F. For each set of mechanical characteristics, average radial strain is calculated at tunnel
face, and results are presented via a chart giving tunnel
face relative deformation (f = uf /R, where uf is extrusion value and R excavation radius) versus security
factor F. Stability is evaluated considering security factor values F(f = 2,5%) and F(f = 5%) giving relative
deformations of respectively f = 2,5% and f = 5%.
Figure 3 shows an example of result chart, where
F(f = 2,5%) = 1.2 and F(f = 5%) 1.4.
Calculation showed that tunnel face stability
depends on tunnel depth. Whatever the formation
considered, both security factors F(f = 2,5%) and
F(f = 5%) were greater than 1 for depth lower than
50 m. For maximal depth (150 m), security factors
where lower than 1. Sensitivity analysis showed that
tunnel face became unstable under 60 to 70 m overburden. Ground behaviour appears to be limit-plastic
at 50 m depth, and totally plastic at 150 m depth.
In order to perform construction using full-face
excavation method, TERRASOL-SETEC TPI advised
face reinforcement would then be necessary.
3.1 Tunnel face stability analysis
3.2
Tunnel face stability analysis was performed using

TERRASOL convergence-confinement code TUNREN. The code uses a C-Phi reduction method, based
on the analytical model EXTRUSION developed by
Wong et al. (1999). Hypothesis are the ones of
convergence-confinement method (circular section,
isotropic, homogeneous and infinite medium, uniform
stress field), with two additional ones:
The construction method chosen was a half-sections

method. Support and lining, as defined in the contract,
were as follows:
PROJECT DESIGN
support: all over the section, including invert, 5 cm

shotcrete confinement layer, plus HEB 180 steel
ribs settled in a 18 cm thick shotcrete layer; steel
ribs spacing had to be defined by design calculations, and was expected to range from 0.75 m to
1.5 m depending on geological conditions;
lining: watertight layer on crown of the tunnel, and
40 to 60 cm thick concrete reinforced at lateral walls
and invert junction; reinforcement sections had to
be defined by design calculations.
tunnel face is supposed spherical, and stress-strain

field follows a spherical symmetry;
tunnel face heading is modelled by taking into
account a decreasing radial pressure at tunnel face.
Material behaviour is supposed elasto-plastic, following Mohr-Coulomb or Tresca criterion.
From initial (c,) short-term values, mechanical
characteristics of the ground are progressively reduced
Support and lining design
In order to estimate tunnel deformations and

solicitations of lining, a staged calculation was
run with finite elements model (FEM). Calculation
387
Table 2. FEM support calculations (after calculation profile name, P and S are respectively for Mainly Pelite and
Mainly sandstones, figure indicates depth).
Crown
ground
settlement
(mm)
Horizontal
convergences (mm)
Calculation
profile
Surface
settlement
(mm)
Ground
Support
P1 (P, 50)
P2 (S, 50)
P3 (S, 150)
P4 (P, 150)
P5 (P, 10)
P6 (P, 30)
3.2
1.1
0.4
0.7
13
4.5
25.8
97.6
1.8
3
44
11
60
190
6.5
8
9
4
16
12
3
3
procedure was lead throughTERRASOL convergenceconfinement code TUNREN, and through French
LCPC finite elements code CESAR. For each profile,
calculation steps were as follows:
Convergence-confinement calculations: estimation
of the deconfined rate 0 after tunnel full-face
excavation, using TUNREN code;
FEM calculations: full-face excavation (deconfinement up to 0 value previously estimated, load
applied to excavation walls set to (N = (1 0 )0 );
completion of support and final deconfinement
(N = 0); completion of lining (mechanical characteristics are set to long-term values, swelling
pressure is applied under invert and ground creeping is taken into account by option EFD of the
calculation code).
Results were expressed as deformations and stress
fields for support, and only as stress fields for lining.
Table 2 sums up displacements results after support
calculations.
Results showed that surface settlements were
expected to be lower than 5 mm.
For a given formation, tunnel deformations were
expected to grow with tunnel depths. For formations
with mainly sandstones, maximal convergences in support and crown settlements are respectively 16 mm
and 25.8 mm, which is rather limited. For mainly
pelitic formations, they are of respectively 12 mm and
97.6 mm, which shows that at maximal depth, tunnel behaviour is completely plastic and deformations
can become important. Numerical model indicates
crown ground settlements about half as high as ground
horizontal convergences for depth greater than 10 m.
Support calculations indicated that great deformations (of about 100 mm) could occur in support at
depth greater than 50 m. Support completion would
then be of a major concern regarding control of those
deformations.
Lining calculations showed that only minimum
reinforcement would be necessary at junction between
lateral walls and invert.
Figure 4. Upper half-section excavation with central part of

the face left in place (Tangier face, around PK 27+960).
CONSTRUCTION
Works began in February 2006, with access cut excavation. Tunnel excavation began in March 2006 at
both portals. Junction of the two faces occurred at PK
28+098 on July the 5th of 2007, after 15 months. Average heading was of about 80 m per month at Tangier
face, and of about 90 m per month at Ras RMel face.
Ras RMel face appeared to be in better geological
conditions than the other face, which allowed higher
heading rate and lighter support: steel ribs could be
spaced from 1,2 to 1,5 m in average, against 0,75 m in
average at Tangier face.
TEC Engineering decided not to apply face reinforcement treatment. As tunnel-face was feared to be
unstable, it was then decided to apply the same halfsections excavation method as used for Meknes tunnel.
In order to control tunnel face stability, during excavation of section upper-half, and in zones with poor
quality ground, central part of the face was left in place
until upper-half support completion (Cf. Fig. 4).
Furthermore, as lower-half excavation and support
completion occurred in average 4 days after upper-half
excavation, behaviour of tunnel face was almost equivalent to a full face excavation, which helped control
deformations.
Opening ranged from 0,75 m in zones with mainly
pelites, to 1,5 m in zones with mainly sandstones, and
was adapted to the deformations measured. A provisional formwork of wooden boards hold by steel bars
prevented shotcrete loss during shotcrete projection,
as can be seen on Figure 3.
A total of 4 crossing zones with enlarged section
was excavated, 2 at each face.
As expected, main concern during construction
was tunnel deformations due to excavation of soft
deformable rocks under high stress level.
388
Surface deformations monitoring and analysis
From each portal, and on a total length of 1500 m, excavation is at a depth lower than 50 m, Ras RMel face
even having to cross a zone at a depth of only 15 m.
Surface settlements where then monitored. Every 20 m
along tunnel axis, a monitoring cross-section was set
using 5 surveyors rods set perpendicularly to the tunnel axis (numbered from 1 to five, with 10 m to 15 m
spacing, central rod number 3 being above tunnel axis).
For a given monitoring section (named following the
corresponding kilometric point of the tunnel), measurements started 2 to 4 weeks before tunnel face
reached the given kilometric point. The frequency was
one measure per day up to stabilisation, and one measure a week once stabilisation reached. Settlements in
the centre part of the tunnel, at greatest depth (150 m),
were not monitored.A chart showing final surface
settlements measured before lining construction is
presented on Figure 2.
4.1.1 Ras RMel face
At Ras RMel face, settlements for a given monitored
section began when tunnel face was approximately
20 m ahead of the corresponding PK. Stabilisation
was reached as tunnel face was 80 m further (approximately after 1 month). At tunnel axis, final settlements
are of about 20 mm for the first 100 m after portal.
Further on, they are inferior to 5 mm. No relation
could be established between tunnel depth and surface
settlements. The low settlements did not allow observation of relevant transverse settlement throughs, and
the crossing of the very low depth zone, between PK
28+924 and PK 28+870, did not induce any increase
of surface settlements.
Surface settlements measured were as expected
after FEM calculation.
4.1.2 Tangier face
At Tangier face, settlements for a given monitored section began as tunnel face was about 20 to 40 m ahead
of the corresponding PK. Stabilisation was reached
when tunnel face was 80 to 120 m further (1 to 1.5
month later). Above tunnel axis, final settlements are
rather high in the first 100 m after the portal (from 20 to
50 mm). These values, higher than the ones calculated,
could be explained by the proximity of the access cut
to the portal, and by the low depth of the tunnel in this
zone. Nevertheless, as a 45 m overburden was reached,
and settlements seemed to decrease, tunnel face hit a
rather weak zone (from PK 26+978 to PK 27+140),
in which surface settlements were much higher than
expected (greater than 20 mm at tunnel axis, reaching
up to 190 mm).
Figure 5 shows settlements evolution at PK
27+000. Tunnel face upper-half reached the corresponding section, when already 20 mm settlements
had occurred at surface. The effect on surface
07/05/06
0
Total settlement (mm)
4.1
21/05/06
4/06/06
18/06/06
02/07/06
Section upper-half
excavation (04/06/2006)
Support completion
(10/06/2006)
-100
-200
Point 1
Point 2
Point 3 (Tunnel axis)
Point 4
Point 5
Figure 5. Settlements at PK 27+000 (Weak zone at Tangier

face).
settlements of tunnel face lower-half reaching the

section, can hardly be seen as no stabilisation had
occurred at that moment. Stabilisation was reached,
two weeks after completion of tunnel support. The
final settlement value above tunnel axis is of 175 mm.
Those settlements can be related to high tunnel deformations as will be discussed further on.
Final settlements for points 5 and 4, respectively
compared with those for points 1 and 2, are lower.
Almost no settlement occurs at point 5. This dissymmetry in transverse settlement through is relevant of
what could be observed at Tangier face: surface terrain is naturally dipping south, so that the cross-section
monitored is in direction of the dip. Points 1 and 2 are
on the upper side of the dip, whereas points 4 and 5 are
on the lower side. Crossing a weak zone exaggerated
this phenomenon.
As no building was constructed above the tunnel,
such high settlements were not of a major concern.
After this zone of high deformations, settlements
are lower than 20 mm. From pk 27+140 to pk 27+340,
surface settlements at tunnel axis increase from 7 mm
to 17 mm, following depth increase from 40 m to
90 m. For greater PK, settlements decrease while depth
increase. As the centre part of the tunnel is reached
(with depth of 150 m), final settlements are of a 5 mm
average, which fits the settlements calculated by the
FEM simulations, and is compatible with measures
relative to Ras RMel face.
Surface settlements were then partly dependent
on tunnel depth, and partly on lithology (greater
settlements in highly deformable soft rock).
4.2 Tunnel deformations monitoring and analysis
As said before, tunnel deformations were expected to
be of a major concern. Every 20 m along tunnel axis,
a monitored cross section was established after support completion, equipped with 5 measuring targets. It
allowed measurements in the support of crown and side
wall settlements, and of section upper and lower-half
convergence.
389
4.2.1 Ras RMel face

From Ras RMel portal, up to junction PK, tunnel
deformations were limited, section upper-half deformations ranging from 10 mm divergence to 30 mm
convergence, crown settlements being of a comparable size. Those values are limited, but higher than
calculated for corresponding profiles (P1, P2, P3, P5
and P6).
From PK 28+620 up to PK 28+400, tunnel deformations are higher, reaching 130 mm convergences
and 78 mm crown settlements. In this zone, crown
settlement is about half of convergences. This local
increase in deformations should be related to decreasing quality of excavated rock rather than to an increase
of overburden. The contact plane between the two flyschs formations might have been crossed along with
this change of rock quality, but could not be identified
firmly.
Approaching PK of faces junction, the deformations
increased and reached values consistent with the ones
measured at the other face.
4.2.2 Tangier face
As shown for surface settlements, after a zone with
rather high deformations due to perturbations induced
by the vicinity of Tangier portal, tunnel suddenly hit
a zone of very high deformations at limited depth
(around 50 m), located between PK 26+978 and PK
27+140.
High convergences in the section upper-half
occurred, greater than 100 mm and reaching 330 mm.
Crown settlements were of values half as high. Trying to secure the zone, it appeared that deformations
stopped as soon as sections support was completed.
Final stabilisation was reached within one month after
section opening, although later creeping could be
identified locally causing support deformations. The
values measured overcome by far the values calculated for support deformations. They are thought to
be linked with local mechanical characteristics for the
rock lower than expected.
22/05/06
Settlement/Convergence value ( mm)
Due to construction constraints, convergence measurements are of questionable reliability. However, it

clearly appears that final measured convergences are
highly variable from one section to another, depending
on lithology rather than on tunnel depth. The convergence values reached range from 20 mm to more than
300 mm, which is far greater than estimated by calculation. Furthermore, differences appear between the
two tunnel faces, Ras RMel face once again showing lower deformations than Tangiers. Figure 2 shows
tunnel deformations.
At both faces, tunnel deformations appeared to stop
as soon as support was completed, closing the excavated sections. Later creeping could be identified as
lining was being constructed.
-40
05/06/06
19/06/06
03/07/06
Crown Settlement (mm)

Section Upper-half convergence (mm)
Section upper-half excavation (23/05/2006)
-140
Support completion ( 30/05/2006)
-240
-340
Figure 6. Greatest tunnel deformations, measured at PK

26+978.
It was feared that deformations would become unacceptable when reaching maximum depth and excavation would become highly unstable. As soon as
it became obvious that support section completion
blocked further deformations, it has been decided to go
on construction with the same method, operating excavation carefully, and keeping opening and steel-ribs
spacing at only 0.75 m as long as necessary.
Figure 6 shows the evolution of tunnel greatest
deformations, which occurred at PK 26+978 (Stabilisation was reached by mid-June).
The high deformations zone goes up to PK 27+780,
where greatest depth is reached. From this PK, and
up to the junction of tunnel faces, deformations are
still important (about 90 to 100 mm in section upperhalf convergences, and 50 mm in crown settlements)
but lower than the previous zone. This is supposed to
be linked with depth increase, and a better stiffness
of soft rock mass. The deformations measured in this
zone, are higher than convergences calculated for the
corresponding profile (P4).
Deformations values increased as junction section
approached, reaching values compatible with the ones
observed at Ras RMel face.
From a general point of view, it is thought that
part of the deformations calculated as ground deformations before support completion (theoretically
non-measurable), are measured as upper-half support
convergences, because of the half-sections excavation
method used, (measures begin as section excavation
is not completed, and upper-half support is not totally
blocked as no invert is constructed at that stage). This
could explain why measured tunnel deformations are
greater that the ones calculated. Section behaviour is
then intermediate between full-face (as calculated in
specific design studies) and half-section excavation.
Face potential instability had been pointed at by
calculation. The cautious excavation method chosen
allowed no collapse to occur on Ras Rmel tunnel
works. The other tunnel for the railway project (SidiAli
tunnel), operated by the same team of company in similar context but with slightly better geological conditions did not receive such cautious treatment at tunnel
390
Design studies gave a rough size range for surface and tunnel deformations, but several zones
with tunnel high deformations were crossed, which
required cautious excavation. Nevertheless, the excavation method used allowed to control deformations,
that were blocked as soon as support was completed.
At time this paper was written, tunnel completion
was programmed for mid-September 2007.
face. Tunnel face collapsed on 31/07/2006, which

shows that TEC Engineering method was appropriate
regarding Ras RMel tunnel face.
4.2.3 Zones with enlarged section
The four zones with enlarged section did not induce
greater deformation compared to surrounding basic
sections.
REFERENCES
CONCLUSION
Ras RMel tunnel had to be excavated in highly variable geological formations, highly deformable soft
rocks, and at a high stress level. To control tunnel
deformations and tunnel face potential instability, it
was then necessary to use half sections excavation
and, when necessary, divided section excavation for
upper-half. This method derived from a method previously used at Meknes tunnel (Marocco), operated by
the same team of companies. During construction, sections behaviour was intermediate between full-section
and half-sections excavations.
Guilloux, A., Le Bissonnais, H. & Pre, M. 2005. Tunnel de

Meknes (Maroc): conception et ralisation dun tunnel en
terrain meuble sous faible couverture, AFTES, Les tunnels, cl dune Europe durable; Proc. Intern. Congress.,
Chambry, 1012 October 2005. Lyon: Spcifique.
Terrasol, Tunren. Tool for tunnel conception (lining and tunnel face stability), convergence-confinement and extrusion calculation code.
Wong, H., Trompille, V. & Dias, D. 1999. Dplacements
du front dun tunnel renforc par boulonnage prenant
en compte le glissement boulon-terrain: approches analytique, numrique et donnes in situ, Revue Franaise de
Gotechnique n 89: 1328.
391
Observed behaviours of deep excavations in sand

B.C.B. Hsiung
Department of Civil Engineering, National Kaohsiung University of Applied Sciences, Kaohsiung, Taiwan, R.O.C.
H.Y. Chuay
Mott MacDonald, Kaohsiung, Taiwan, R.O.C.
ABSTRACT: In this paper, structural and ground behaviours of several excavations in Kaohsiung, Taiwan
were described and examined. Based on behaviours observed from these excavations, it was found that the
maximum lateral wall displacement (hmax ) in relation to the maximum excavation depth (He ) is approximately
0.034 to 0.3%. The ratio of maximum surface settlement (vmax ) to hmax varies from 0.5 to 0.7 for the excavation
constructed by bottom-up method and from 1.3 to 1.8 for the excavation using a semi-top-down method. The zone
at ground level affected by the excavation is up to 3 times the maximum excavation depth behind the diaphragm
wall which retains the earth during construction. This finding is different from previous conclusions reported
by Clough & ORourke. It is apparent that chemical churning pile does not effectively reduce influences of the
excavation on adjacent buildings.
INTRODUCTION
Table 1.
Ground deformations induced by deep excavations in

clays have been explored widely (Wong et al., 1997,
Hsieh & Ou, 1998, Hsiung, 2002, Liu et al., 2005) but
studies regarding observed behaviours of deep excavations in sand are comparatively limited (Burchell, 2000
and El-Nahhas, 2006). In this paper, case histories
from excavations in Kaohsiung City, Taiwan provide an opportunity to explore structural and ground
behaviours induced by excavations in sand. Empirical approaches for evaluating lateral wall movements
and surface settlements were studied and discussed.
Further, the effectiveness of chemical churning pile
for house protection and prediction of prop load were
also discussed in this study.
Construction activity
Excavate to 3.4 m below ground level
Install 1st level prop at 2.5 m below
ground level
Install 2nd level prop at 5.9 m below
ground level
Install 3rd level prop at 9.1 m below
ground level
Install 4th level prop at 12.6 m below
ground level
ground level
2 THE SITES
Several deep excavations for the Orange Line of
the Kaohsiung Rapid Transit Systems (KRTS) have
recently been carried out in an area where the ground
conditions are relatively uniform. Among them, excavations of three underground stations, O6, O7 and O8
were selected for this study. The sites are located at
the centre of Kaohsiung City, Taiwan and the maximum excavation depth varies from 19.6 m to 20.9 m.
Lengths and widths of these excavations are in the
range of 178 to 215 m and 22 to 24 m, respectively. All
excavations are retained by 1.0 m thick reinforcement
concrete diaphragm walls. Construction sequences of
Construction sequence at O6.

Period
(day/month/year)
24/10/0327/10/03
02/11/0306/11/03
08/11/0312/11/03
13/11/0316/11/03
24/11/0327/11/03
04/12/0307/12/03
14/12/0316/12/03
28/12/0331/12/03
06/01/0408/01/04
09/01/0412/01/04
25/01/0402/03/04
O6, O7 and O8 are listed in Tables 1 to 3. H-type

steel sections were selected for horizontal props and
the levels of props at these sites are also described in
Tables 1 to 3. The horizontal spacing of props is similar at O6. O7 and O8 and is approximately 4.5 m. In
addition, angle bracing and waling were constructed
using H-type steel sections in order to strengthen the
strut system.
The main soil stratum of these sites is similar and
consists of medium to dense silty sand with several
393
Table 2.
Period
(day/month/year)
09/12/0313/12/03
26/12/0303/01/04
Silty Sand
5
Silty Clay
10
08/01/0417/02/04
18/02/0422/02/04
Silty Sand
06/03/0408/03/04
09/03/0414/03/04
Excavation
depth 19.57m
15
20
15/04/0418/04/04
19/04/0421/04/04
Silty Clay
Silty Sand
22/04/0426/04/04
27/04/0429/04/04
Silty Clay
1m thick ,
36m deep
diaphragm
wall
25
30
Silty Sand
31/05/0431/05/04
Depth(m)

ground level
ground level
ground level
ground level
ground level
Lateral wall displacement(mm)

-5
15
35
55
35
40
Table 3.
27/02/0402/03/04
03/03/0407/03/04
18/03/0422/03/04
23/03/0427/03/04
-8
Silty Sand
52
72
0
10
Silty Sand
Excavation
depth 21.7m
Silty Clay
1m thick ,
38m deep
diaphragm
wall
14/08/0416/08/04
17/08/0419/08/04
08/09/0415/09/04
15/09/0418/09/04
32
Silty Clay
30/03/0402/04/04
03/04/0307/04/04
08/04/0410/04/04
11/04/0415/05/04
12
Silty Sand
15
20
25
Depth(m)

ground level
ground level
ground level
Construct concourse-level slab (0.8 m
thick) at 14.2 m below ground level
ground level
ground level
Figure 1. Lateral wall deformations and ground profile

at O6.
Period
(day/month/year)
30
35
19/09/0422/09/04
40
thin layers of clay. The SPT-N value of ground is

from 5 to 30. The initial groundwater level was 36 m
below ground level and it remained hydrostatic before
commencement of excavation. Most excavations are
surrounded by 4-storey to 7-storey buildings but some
high-rise buildings (up to 12-stories) are also observed
on site.
3
OBSERVATIONS
Monitoring instruments installed on site included

inclinometers inside the diaphragm wall and soils,

at O7.
observation wells, standpipe/electrical piezometers,

vibration wire gauge on struts, tiltmeter and prcision
levelling on buildings.
Figures 1 to 3 show the lateral wall deformations
observed from O6, O7 and O8 at the final excavation
stage together with their ground profiles, respectively.
It is evident that the maximum lateral wall movement
varies from 32 to 64mm. The shape of lateral wall
deformation is a cantilever at the shallow excavation
394
Settlement(mm)

-5
5
15 25 35
0
5
Silty Sand
10
15
20
Silty Clay
Silty Clay
25

at O8.
Distance to the wall(m)
Settlement(mm)
10
20
30
40
50
60
70
End of excavation
1/4/04
4/3/04
7/2/04
ELP001 31mbgl
ELP002 36mbgl
ELP003 31mbgl
10
End of excavation
15
20
25
70
30
10
Figure 7. Piezometric levels inside the excavation at O7.
15
20
25
Shallow excavation
30
End of excavation
Measured date
05/25/03 08/23/03 11/21/03 02/19/04 05/19/04 08/17/04
0
10

20
30
40
Piezometric level(m,bgl
Figure 4. Surface settlement at O6.
Settlement(mm)
60
Shallow excavation
Start of excavation
35
40

20
30
40
50
Measured date
10/6/03
0
30
Silty Sand
10
Piezometric level(m,bgl)
Silty Sand
1m thick,
37m deep
diaphragm
wall
Depth(m)
Excavation
depth 20.9m
0
5
10
15
20
25
30
50
5
10
15
20
1
2
Start of excavation
End of excavation
PS005 27m bgl
PS006 34m bgl
PS007 27m bgl
PS008 34m bgl
4
5
6
7
25
Shallow excavation
30
End of excavation
Figure 8. Piezometric levels outside the excavation at O6.
level but tends to become prop-mode the deeper the

excavation. The depth where the maximum lateral wall
movement occurs is generally at the same depth of
excavation.
Figures 4 to 6 present induced surface settlement
and the observed maximum surface settlement is up to
20 to 28 mm at the completion of excavation. However,
as shown in Figures 4 to 6, limited data was available

for the area close to the excavation.
Since piezometers inside the excavation at O8
were all broken, only piezometric levels inside the
excavation at O6 and O7 were measured, as indicated in Figure 7. The piezometric levels continue to
decrease as excavation progresses. Figure 8 indicates
the piezometeric level outside the excavation. It is
395
Figure 9. The relationship of vmax , hmax and He
Figure 10. The relationship of the depth of the maximum

lateral wall displacement (Dm ) and excavation depth (He ).
apparent that it is affected by a change of pore pressure

inside the excavation therefore it shows a slight drop
as excavation progresses.
The loads on struts were also measured and the maximum observed measured load is in the range of 220
to 320 tons.
4.2 Surface settlement troughs
4
4.1
DISCUSSIONS
Maximum movements
Figure 9 presents the relationship of maximum lateral

wall movements and excavation depth at O6, O7 and
O8. It is found that the maximum lateral wall displacement (hmax ) in relation to the maximum excavation
depth (He ) is approximately 0.034 to 0.3%. The ratio
of hmax /He is comparatively high at O7 and this could
be accredited to less stiffness of the supporting system. Mana & Clough (1981) described the ratio of
hmax /He mainly in the range of 0.4 to 2.0%. Ou et al.
(1993) reported that the same ratio was approximately
from 0.2 to 0.5% for excavations in Taipei. The ratio
obtained from this study tends to be lower and it is
considered that varying ground conditions could be
the reason for the difference.
As shown in Figure 10, the depth where maximum
lateral wall movement occurs is in a range of 0.86 to
1.26He . The maximum lateral wall movement is close
to the excavation level.
The lateral wall movement at O8 is smaller than the
others and it is assumed a semi-top-down construction
method might provide a greater strut stiffness and
reduce lateral wall movement in the excavation.
As shown in Figure 9, the maximum surface settlement

(vmax ) is approximately 0.050.13% of He . Wang
(2003) suggested that vmax induced by the excavation in Kaohsiung might reach 0.04% to 0.25% of He .
Observations in this study appear to be consistent with
Wangs conclusion.
Further, the relationship between vmax and hmax
was explored and Figure 9 presents the distribution of
the calculated ratio of vmax to hmax . It appears that
the ratio of vmax /hmax falls in a range of 0.3 to 0.75
at O6 and O7 but tends to be greater at O8 (approximately from 1.3 to 1.8). Since different construction
methods (bottom-up for O6 and O7 but semi- topdown for O8) were employed this is considered to be
the main reason for the difference. As additional lateral wall movements at O6 and O7 were generated
by the delayed installation of struts at several stages,
this might reduce the ratio of vmax to hmax. Some
data from O6 and O7 were therefore ignored, hence
it is suggested that the ratio of vmax /hmax varies
from 0.5 to 0.7 at O6 and O7. However, only limited data were collected for areas near the excavation
(0 to 5 m from the excavation). This might affect
the accuracy of the measurement of ground surface
settlement.
Clough & ORourke (1990) suggested that the
influence zone at surface level affected by the excavation in sand was two times the excavation depth
for the ground behind the wall retaining the soil,
but it tends to be greater (up to three times the
excavation depth) at O6, O7 and O8. Pumping from
deeper levels under ground is thought to be the reason
for that.
396
6/18/03
0
9/16/03
Measured date
12/15/03
3/14/04
Strut load(kPa)
6/12/04
9/10/04
50
100
SB69
SB70
Depth(m)
End of excavation
10
SB71
Settlement(mm)
SB81
20
SB87
150
200
250
Peck
Field measurement
5
10
15
30
40
20
Start of excavation
Strut load(kPa)
0
50
50
100
150
200
250
0
Installation of CCPs
Depth(m)
60
Figure 11. Building settlement at O6.
Twine and Roscoe

Field measurement
5
10
15
Use of chemical churning pile
20
Figure 12. Comparison of prop load from field measurement and apparent earth pressure at O6.
Strut load(kPa)
0
50
100
150
200
250
Peck
Field measurement
5
Depth(m)
Chemical churning piles (CCP) are commonly used in

Taiwan in order to reduce the influence on buildings
resulting from adjacent excavation. Materials used for
CCP generally include cement and water. Woo & Moh
(1990) reported that the movements of the building
could be prevented if the tip of the chemical churning pile was installed deeper than the potential failure
surface. CCPs were used at O6 in order to protect structures nearby. For the design of CCPs, piles have to
penetrate at least 2 meters below the 45 active failure surface of the excavation (Maeda- Longda Joint
Venture, 2003). CCPs were installed between the excavation and buildings, approximately 7.59.0 m from
the excavation. The diameter of the pile is 0.35 m and
the length of the pile varies from 12.0 to 14.5 m. During the construction, pressure used for jet-grouting was
kept at 19.6 MPa. The rate of rotation and raising of the
rod were 20 rpm and 4.0 min/m, respectively.Although
CCPs were installed around O6, the buildings adjacent
to O6 continued to settle after installation of CCPs,
as shown in Figure 11. CCPs may only provide very
limited influence on reducing continual settlement of
buildings and this is consistent with observations that
Hsiung (2002) reported. Insufficient CCPs stiffness
could be a possible reason.
10
15
20
25
Strut load(kPa)
0
5
Depth(m)
4.3
50
100
150
200
250
Twine and Roscoe

Field measurement
10
15
20
25
Figure 13. Comparison of prop load from field measurement and apparent earth pressure at O7.
4.4 Prop load

Peck (1969) suggested that the prop load could be
estimated associated with the apparent earth pressure
method. Consideration of case histories from different
projects, Twine & Roscoe (1997) revised the apparent earth pressure method. Figures 12 and 13 present
the comparison of strut load from field measurement
and apparent earth pressure suggested by Peck and

Twine and Roscoe. It suggests that Pecks method obviously underestimates the actual load on the prop but
the method suggested by Twine and Roscoe provides a
397
more accurate prediction in practice. Such conclusion

is similar to those by Ou et al. (1998) and Hsiung
(2002).
CONCLUSIONS
This study provides an opportunity for exploration

of ground behaviours induced by excavation in sand.
Based on this study, conclusions can be made, as
follows:
1. The lateral wall movement induced by excavations
selected in this study is up to 0.3% of the excavation depth. Such displacement is less than those
found from several related previous literatures. It is
suspected the varying ground conditions could be
a factor in the difference.
2. The maximum lateral wall displacement is 0.86 to
1.26 of the excavation depth. The depth where the
maximum lateral wall movement occurs tends to be
close to the excavation level
3. Associated with observations in this study, the ratio
of vamx to hmax varies from 0.5 to 1.8 and the
use of construction method (bottom-up and semitop-down) might induce the obvious change in this
ratio.
4. The influence zone caused by excavation can be up
to three times the excavation depth.
5. The installation of CCPs does not provide an
effective solution for reducing the settlement of
buildings near the excavation. Insufficient CCPs
stiffness might be a reason.
6. The apparent earth method suggested by Peck
underestimates the prop load. The method suggested by Twine and Roscoe might provide a more
acceptable result in practice.
ACKNOWLEDGEMENTS
The authors would like to thank for the financial support from National Science Council, Taiwan for this
project (Project Number: NSC92-2218-E-151-003).
Data for this research provided by Maeda- Longda
Joint Venture and comments on paper content and
English from Mr. Tony Shield are also appreciated.
REFERENCES
Burchell, A. J. 2000, Cairo metro line 2- construction problems and their solutions,TunnellingAsia2000, New Delhi
Clough, G. W. and ORourke, T. D. 1990, Constructioninduced movement of insitu walls, Design and performance of earth retaining structure, ASCE, special
publication, No. 25, pp. 439470
El-Nahhas, F. M. 2006, Tunneling and supported deep excavations in the Greater Cairo, International symposium on
utilization of underground space in urban area, Sharm
El-Sheikh, Egypt
Hsieh, P. G. and Ou, C. Y. 1998, Shape of ground surface settlement profiles caused by excavation, Canadian
Geotechnical Journal, Volume 35, No. 6, pp. 10041017
Hsiung, B. C. 2002, Engineering performance of deep
excavations in Taipei, PhD thesis, University of Bristol
Liu, G. B., Ng, C. W. W. and Wang, Z. W., 2005, Observed performance of a deep multistrutted excavation in Shanghai
Soft Clays, Journal of Geotechnical and Geoenvironmental Engineering., volume 131, issue 8, pp. 10041013
Maeda- Longda Joint Venture, 2003, Working proposal for
house protection at O6 station (in Chinese)
Mana, A. I. and Clough, G. W. 1981, Prediction of movements for braced cut in clay, Journal of the Geotechnical
Division, Proceeding of the American Society of Civil
Engineers Volume 107, No. GT6, pp. 759777
Ou, C. Y., Hsieh, P. G. and Chiou, D. C. 1993, Characteristics
of ground surface settlement during excavation, Canadian
Geotechnical Journal, Volume 30, pp. 758767
Ou, C. Y., Liao, L. T. and Lin, H. D. 1998, Performance of
Diaphragm Wall Constructed Using Top-Down Method,
Journal of Geotechnical and Geoenvironmental Engineering, volume 124, issue 9, pp. 798808
Peck, R. B. 1969, Deep excavation and tunneling in soft
ground, proceeding of the 7th international conference
on soil mechanics and foundation engineering. Mexico
City, state of art volume, pp. 266277
Twine, D. and Roscoe, H., 1997, Prop loads guidance on
design, Funders Report FR/CP/48, 224 pages, CIRIA,
London
Wang, C. C. 2003, the study of engineering characteristics
of deep excavations in Kaohsiung, NSC research report,
Taiwan (in Chinese)
Wong, I. H., Poh, T. Y. and Chuah, H. L. 1997, Performance of Excavations for Depressed Expressway in Singapore, Journal of Geotechnical and Geoenvironmental
Engineering, Volume 123, Issue 7, pp. 617625
Woo, S. M. and Moh, Z. C. 1990, Geotechnical characteristics
of soil in Taipei basin, Proceedings of the 10th Southeast Asian Geotechnical Conference, Volume 3, Taipei,
pp. 5165
398
Environmental problems of groundwater around the longest expressway

tunnel in Korea
S.M. Kim, H.Y. Yang & S.G. Yoon
Bau Consultant CO. LTD., Seoul, Korea
ABSTRACT: Long tunnels are usually located in mountainous areas with limited infrastructure. Therefore
very little is known about the geological, hydrogeological and geotechnical conditions: the longer the tunnel, the
higher the probability of encountering adverse conditions for tunneling ; the greater the cost and duration for
tunnel construction. Inje tunnel will be the longest expressway tunnel in Korea. In case of Inje tunnel, Special
problems related to the length of the tunnel are for example typically the logistics, ventilation and environmental
impacts. Even though modern excavation methods of tunnels have been developed, various types of problems
such as change in groundwater distribution and transformation of geographical features still remain. It is not
uncommon that private wells and small streams are used for daily life in the regions where mountain tunnels are
located. Then serious social problems such as well water level fall, being attributable to tunnel excavation occurs.
In the design stage, we evaluated that the quantity of leakage water into tunnels and groundwater drawdown
area was simulated using numerical modeling such as MODFLOW and MAFIC to reduce adverse effects on life
environment around tunnels.
INTRODUCTION
Nowadays road construction projects in Korea have

shown a tendency toward linear route designs, for
the purpose of increasing running speed and reducing transportation time. Moreover, a growing number
of projects are designed with tunnels and bridges in
order to minimize damage to the natural environments,
which can be caused by slope cutting.
However, some road construction projects, which
were started after the year 2000, provoked the delay
of projects and creating a lawsuit battle. Such a problem precipitated because the impacts of groundwater
outflow during tunnel excavation were dealt with
superficial experiences rather than scientific verification. Since 1998, there have been four important
environmental conflicts related to tunnel groundwater outflow in Korea. Every tunnel construction that
was mentioned above has passed legal environmental impact assessment (EIA) procedures. Nevertheless,
these construction projects were delayed or stopped,
within a span of 2 or 3 years on average, due to
strong objections from civil environmental NGOs and
local residents near the project sites. The NGO insisted
that the groundwater outflow during tunnel excavation
accompanied more serious secondary environmental
impacts such as abrupt changes 4 Mitigation Plan for
Environmental Impacts of Groundwater in Tunnels in

the fauna and flora ecology near the upper part of the
tunnel sites. (Lee, J. et al 2005).
Generally speaking, the movement of groundwater
is based on meteorology conditions, surface vegetation, hydrogeological conditions and so on. Because a
tunnel is a linear structure, the possibility of changing
hydrogeology conditions is high and it is difficult to
avoid adverse effects to life environment in regions
where small-scale water usage remains. Therefore,
many studies on the quantity of leakage water into tunnels and groundwater drawdown area are executed to
reduce adverse effects on life environment around tunnels as well as retain safety and workability in tunnel
constructions.
In this study, we performed the following investigation and numerical analysis to evaluate the environmental influence of groundwater with excavation of
tunnel in mountainous area;
399
Ground Investigation: Borehole logging, geophysical survey, BIPS, Lugeon test, groundwater level
monitoring, water well survey and so forth
Numerical Analysis for groundwater flow: Continuous and fracture media model
Evaluation of influence: drawdown of groundwater
level, groundwater inflow rate into tunnel Etc.
OVERVIEW OF THE TUNNEL
The longest tunnel in Korea will be the Inje tunnel

with the length of 11 km. The tunnel consists of two
parallel double-lane tube each with a width of 14.5 m.
The tubes are connected by cross tunnel as emergency
facility every 250 m and 750 m. The gradient of the
tunnel from Chuncheon to Yangyang is 1.95% downward. An Incline tunnel with the length of 1.5 km was
designed on the purpose of access for excavation face
and escape tunnel in emergency. Table 1 shows general
overview of the tunnel.
3
3.1
GROUND INVESTIGATION
General overview on the topology and
geological conditions
The dominant topographic features of the area are

rugged mountains and a few of streams. The western
part of the area is lower altitude and more relief than
the eastern part. This area shows early mature stage in
geomorphologic cycle.
This area consists of Pre-Cambrian porphyroblastic gneiss, banded gneiss, and Jurassic biotite granite,
two-mica granite, Cretaceous basic and acidic dikes,
and Quaternary alluvium and diluvium as shown in
Figure 3.
3.2
NUMERICAL ANALYSIS
4.1 General aspects of groundwater flow model

The environmental assessment of tunnel groundwater aims to forecast the drawdown and variation of
domestic groundwater near planned routes due to
tunnel excavation. As usual, numerical methods of
groundwater flow modeling are used for the environmental impact assessment of tunnel groundwater.
The program packages used for the modeling of
tunnel groundwater outflow are different from each
other, depending on the regional groundwater level
variation in whole tunnels or small-scale groundwater level variation in narrow fracture zones. Normally,
continuous numerical model packages are commonly
used for modeling large-scale groundwater flow variations, while fracture media models are applied for
Table 1.
Layouts of Inje Tunnel.
Length
Alignment
Gradient
Shape of Portal
Traffic Type
Ventilation
Excavation Method
Emergency Facilities
Ground investigation
We carried out geological and geotechnical survey as

shown in Figure 2.
Figure 1. Location of Inje tunnel and its vertical section profile.
400
10,965 m
R = 2000 R = 4991.3
1.95%
Arch wall type
2 tube (2- lane)
Jet Fan : 80
Ventilation Shaft : 4
Drill and Blast
Escape connecting tunnel : 44
Interval : 750 m for vehicle
: 250 m for human
groundwater flow modeling in smaller zones composed of jointed rock aquifers.

Similarly, when we carry out the modeling of
groundwater outflow, the MODFLOW package is
commonly used for numerical analysis of regional
groundwater flows in whole areas of tunnel and
drainage, and MAFIC is applied for detailed groundwater modeling in small discrete zones in tunnel
areas.
In this study, two representative models are used
to forecast the tunnel groundwater outflow pattern as
following Table 2.
The ground conditions of modeling section are
Very-good(rock type 1) Moderate(rock type 3) grade
in rock classification and the rock covers of tunnel

under stream and valley area are about 100 m or higher.
4.2 Continuous media modeling

4.2.1 Setup of the model
MODFLOW was used to simulate groundwater flow
as continuous media model in the studied area. The
whole area is 4 km 3.5 km.
The boundary of the model is set at the summit
of northern mountains and southern mountains in the
upper and lower end, respectively.
Hydraulic conductivities are classified into fifteen groups according to the permeability of the
regions from Lugeon test. The range of value is
4.5E-5 cm/sec 7E-6 cm/sec.
4.2.2 Steady state simulation
We calibrated the value of groundwater head between
the calculated and observed one in order to modify
Table 2.
Program
MODFLOW
MAFIC
Model
3D-Continuous
Model
Sta.
2 + 500 6 + 500
3 D-Discontinuous
Model
Sta.
4 + 000 4 + 500
Section
Figure 2. Flow chart of Ground Investigation.
Layout of groundwater flow modeling.
Figure 4. Ground condition and modeling section.
Figure 3. Geological features of Inje tunnel.
401
Figure 7. Distribution of equivalent head in transient state

condition.
Figure 5. Boundary condition of the model.
Figure 6. Distribution of equivalent head in steady state

condition.
the hydraulic model. Figure 6 shows distribution of

equivalent head in steady state condition.
We can see that groundwater inflow was concentrated at tunnel alignment and flow from the summit
areas of northern and southern mountain to Bangtae
stream.
4.2.3 Transient state simulation
Result of Transient flow simulation is used to evaluate
variation of groundwater level with tunnel excavation.
We can know that the change of groundwater level
with time is very small amount of 0.008 m 0.015 m
before tunnel excavation.
4.2.4 Change of groundwater level due to tunnel
excavation
We can see the distribution of groundwater level and
drawdown at the check point due to the tunnel excavation as following Figure 8 andTable 3.The check points
in model are located every 500 m from Sta. 3 + 000 to
Sta. 6 + 500.
The drawdown of groundwater due to tunnel excavation is 1.92 m 3.29 m as you can see the Figure 8.
It is small amount in general, but we worried about
the influence on groundwater systems around tunnel.
Therefore, we carried out simulation for waterproof
Figure 8. Distribution of groundwater level.
grouting with the aim of reducing water inflow into

tunnel in rock mass.
Groundwater level at Sta. 5 + 000 will fall down
to 2.03 m after tunnel excavation and water inflow
rate per km is 0.127 m3 /min. In case of using waterproof grouting to the ground near tunnel, drawdown
of groundwater level is 0.79 m and water inflow rate
402
Table 3.
tunnel.
Groundwater level and water inflow rate into the
Check point
Drawdown of
groundwater level
(m)
(without grouting)
Drawdown of
groundwater level
(m)
(with grouting)
Sta. 3 + 000
Sta. 3 + 500
Sta. 4 + 000
Sta. 4 + 500
Sta. 5 + 000
Sta. 5 + 500
Sta. 6 + 000
Sta. 6 + 500
3.20
2.94
2.32
1.92
2.03
2.98
3.20
3.29
0.94
0.87
0.58
0.73
0.79
1.04
0.77
1.05
Water inflow
rate per km
0.127 m3 /min
(730.75 m3 /day)
0.060 m3 /min
(343.27 m3 /day)
Table 4.
Major joint set and fault in this area.
Joint
set
Orientation
Set 1
Set 2
Set 3
Fault
80/194
78/264
30/124
85/105
Length
(m)
S.D of
Length
(m)
Distribution
of length
15.36
12.93
34.47
10.38
9.63
13.18
Lognormal
Lognormal
Lognormal
Figure 10. three dimensional model of fracture network and

tunnel for steady state simulation.
Figure 9. Drawdown groundwater level in case of applying

waterproof grouting at Sta. 5 + 000.
per km is 0.060 m3 /min. It can be notified that if

waterproof grouting into the ground is performed in
advance, the amount of groundwater inflow into the
tunnel will decrease as shown in Figure 9 and Table 3.
4.3 Fracture media modeling
MAFIC (Matrix And Fracture Interaction Code) determines solutions of flow and pressure by conjugate
gradient finite element methods, and solves solute
transport by particle tracking discrete fracture network
models. Finite element meshes are based on fracture
network analysis and geometric modeling.
4.3.1 Set up of the fracture model
Using the package of MAFIC, we simulated groundwater flow through fracture model in the studied area
(Sta. 4 + 000 4 + 500). Input data was estimated
based on joint data from detailed surface and borehole survey. In facture model, three of major joint set
was estimated as following Table 4.
Figure 10 shows three dimensional fracture network
and tunnel.
4.3.2 Steady state condition
As the stage of setting boundary condition in steady
state simulation, we estimated groundwater level and
Figure 11. Distribution of water head in steady-state condition before tunnel excavation.
transmissivity(T) based on results of MODFLOW in

steady-state flow condition, water well survey and
borehole test and so on.
Especially, average transmissivity evaluated from
the result of Legeon test and BIPS is 1.24 m2 /sec.
403
CONCLUSIONS
In spite of the difficulty for estimation of water flow

characteristics into tunnel before excavation, it is very
important to predict the variation of hydraulic system
due to the tunnel.
In this study, two representative models are used to
forecast the tunnel groundwater outflow pattern such
as MODFLOW and MAFIC.
Based on the result of numerical modeling, it can
be notified that if waterproof grouting into the ground
is performed in advance, the amount of groundwater
inflow into the tunnel will decrease compared to nongrouting case.
REFERENCES
Figure 12. Distribution of water head in transient-state
condition before tunnel excavation.
Figure 11 shows distribution of water head in

steady-state condition before tunnel excavation. It was
set as the value of initial water head in transient flow
condition.
4.3.3 Transient state condition
We simulated the groundwater flow due to tunnel
excavation in the condition of transient state. As tunnel excavation was performed, groundwater flowed
into tunnel and water head fell down suddenly and
afterward gradually.
In the case of 500 m tunnel advance, water inflow
rate per km into tunnel is 0.10 m3 /min.That was similar
to the result of MODFLOW (0.127 m3 /min).
Kim, S. & Yoon, S. 2006a. Environmental Problems of

Groundwater around the mountain tunnel (2),TunnelTechnology Journal of KoreanTunnelAssociation,Vol. 8, No. 2,
8298.
Kim, S. & Yoon, S. 2006b. Environmental Problems of
Groundwater around the mountain tunnel (1),TunnelTechnology Journal of KoreanTunnelAssociation,Vol. 8, No. 1,
6478.
Kim, S., Yoon, S. & Yoon, J. 2005. Influence of Groundwater
in the Railway Tunnel Construction Korea Society for Rock
Mechanics 2005 Fall National Conference, October 27,
2005, Chuncheon, Korea.
Lee, J. & Lee, Y. N. 2005. A Study on Mitigation Plan for
Environmental Impacts of Groundwater in Tunnels. Korea
Environment Institute, 536.
404
Measurements of ground deformations behind braced excavations

T. Konda
H. Ota & T. Yanagawa

Osaka Municipal Transportation Bureau, Osaka, Japan
A. Hashimoto
Kotsu Service Co., Ltd., Osaka, Japan
ABSTRACT: The deformation behavior of earth retaining wall and backside ground due to braced excavation
were measured at the construction sites of Osaka Subway Line No. 8. It was observed that backside ground
moved to excavation & down side along a circular slip following the wall deflection, and its influence spread
to the ground surface. The backside ground surface settlement near the wall was suppressed due to the friction
between wall and surrounding ground, and maximum settlement occurred at some distance from wall in the
4560 degree area from the bottom of the wall. The backside ground deformation area decreased exponentially
with the distance from the wall. As a result each observed data from monitoring sites was analyzed statistically
according to the soil characteristic, wall deflection and surface settlement in prominent soft clay layer were larger
than those in prominent sand and gravel layer. The relation between the wall deflection area A and backside
ground settlement area As was that As
= (0.20.3) A in prominent soft and sensitive clayey layer. The backside
ground settlement of this observed data converged smaller than past data.
INTRODUCTION
There are many neighboring constructions in recent

subway works. In order to guarantee the construction
safety and control the influence on the surrounding
environment to the minimum, it is necessary to use
the observational method and prediction results effectively. For example, in the case of cut and cover method,
which is usually used to construct the subway station,
it is necessary to evaluate the backside ground deformation of earth retaining wall and groundwater level
fluctuation, to carefully monitor some particular concern points, and of course to estimate the safety of
timbering of a cut and the excavation bottom.
In this paper, earth retaining wall deflection and
backside ground deformation due to braced excavation
were evaluated circumstantially in open cut construction sites, which main excavation soil was the alluvial
soft clay, in Osaka Subway Line No. 8. All observed
data in 11 monitoring sites were got together and analyzed with the same idea. The evaluated results were
described for each soil characteristic. Some behaviors
of ground deformation due to braced excavation based
on these monitoring results were considered.
CONSTRUCTION SUMMARY
Osaka Subway Line No. 8 was constructed at the east

side of Osaka city as northsouth rail route in underground. Every station was built by open cut method.
In order to control the settlement at the ground surface
and the influence on the neighboring ground environment, seepage control method was adopted in all of the
station construction sites with extending earth retaining walls, where the most were Soil Mixing Walls, to
the Pleistocene clayey layer under the artesian aquifer.
It is possible to classify three areas roughly based on
the difference of soil characteristic as Figure 1 shows.
This new rail route located at the East side of Uemachiplateau, and this area was an inland sea in ancient
times.
(1) North Area
This area interleaved the Yodo-river, and constitutes
the alluvial layer, upper and lower Pleistocene. Alluvial layer is composed of a fine sandy layer having
low uniformity coefficient Uc, and soft and sensitive
clayey layer having unconfined compressive strength
qu = 40100 kN/m2 . Tenma-gravel layer accumulated
405
Figure 1. Plain view and soil crossing view of Osaka Subway Line No. 8 (Unei et al., 2002).
continuously with just about uniform thickness as

upper Pleistocene layer with high permeability. Lower
Pleistocene layer Osaka Group is laminated ground
with stiff clayey layer of cohesion c = 300600 kN/m2 ,
and dense sandy layer of Standard Penetration Test
N-value is more than 60. Each Osaka Group dips in
the south and east direction neighbor the Yodo-river.
(2) Middle Area
This area is constituted of the thin alluvial layer, upper
and lower Pleistocene. Upper Pleistocene layer is constituted with stiff clayey layer (N-value = 510) and
very dense sand and gravel layer with high permeability. Lower Pleistocene layer Osaka Group is laminated
ground with stiff clayey layer (c = 200400 kN/m2 )
and dense sandy layer (N-value > 60). The ground of
this area is harder than other areas comparatively.
(3) South Area
This area located in Neyagawa-lowland and it is different from the rest area for that the very soft and sensitive
alluvial clayey layer (N-value = about zero, qu = 40
100 kN/m2 , liquid limit IL = about 1.0) deposited
within about 15 meters. Under the Alluvial layer, upper
Pleistocene layer (N-value = 2260) with high permeability and lower Pleistocene layer Osaka Group,
with stiff clayey layer (c = 200400 kN/m2 ) and dense
and slightly cohesive sandy layer (N-value = 30 and
above 60). Each Osaka Group dips in the south and

east direction parallel to Uemachi-plateau.
3
BEHAVIOR OF BACK GROUND

DEFORMATION DUE TO BRACED
EXCAVATION
Here, some monitoring results of the earth retaining

wall deflection and backside ground deformation due
to braced excavation at the A-site are represented and
the dispersed behavior in backside ground from wall
to ground surface is described as Figure 2 shows (Ito
et al., 2006).
Alluvial layer and upper & lower Pleistocene layer
are formatted from the ground surface near the A-site.
Alluvial layer is composed of a fine sandy layer (1st
aquifer) with 2 m thickness having low Uc and N-value
equals to about 2, and a soft and sensitive clayey layer
with N-value = 03, IL = 0.41.0, c = 20100 kN/m2 .
This Alluvial clay layer is a typical soft layer in this
construction site. On the other hand, upper Pleistocene
sandy and gravel layer Tsg (2nd aquifer) with partially scattered gravel, lower Pleistocene clayey layer
Oc3 (c = about 400 kN/m2 ) and lower Pleistocene
sandy layer Os3 (3rd aquifer, N-value >60) existed
continuously under the Alluvial layer.
Seepage control method was adopted in this
construction site with extending the wall to the low
406
Figure 2. Earth retaining wall deflection, backside ground deformation vector and distribution of backside ground surface
settlement at the A-site.
permeable layer Oc3 (GL-26.3 m). The excavation

width was 16.2 m, and the final excavation depth was
GL-21.5 m, in short, the wall penetration depth is
4.8 m. Because the soil improvement work was carried
out near the monitoring site after GL-1.5 m excavation depth to protect the deficient part of wall, initial
value of each observed data was set the value after 1st
excavation to GL-4.5 m.
The maximum wall deflection occurred at a slightly
deeper depth from the excavation bottom due to the
braced excavation till the 6th excavation step, and this
influences reached the backside ground and surface.
For example, the maximum wall deflection was 38 mm
at the excavation bottom, which was the middle ofAmc
layer (GL-12.5 m), in the 5th step. According to this
influences, the maximum horizontal ground deformation of the No. 1 was about 29 mm at the slightly above
the maximum deformation depth of wall. The maximum horizontal ground deformation of the No. 2 was
about 12 mm in Auc layer and Aus layer. It was known
that the wall deflection caused the back side ground
deformation like the circular slip and this influence
reached the ground surface.
On the other hands, the maximum surface
settlement at the No. 2 was 6 mm, and the settlement
distribution was convex downward with 0 mm settlement at the point, 20 m outside from the wall.
It was thought that the friction between the wall
and surrounding ground supported the surface settlement. These ground movements happened in
the 4560 degree area from the bottom of the
wall.
The maximum wall deflection increased to 51 mm
at just above from the excavation bottom, which is
the lower of Amc layer (GL-14 m), in the 6th step.
Since the 6th step, wall deflection did not increase any
more. It was believed that the cause of this inhabitation
phenomenon was from the subgrade reaction in theTsg
and Oc3 layer was large and the effects of pre-load on
the wall.
Figure 3 shows the definition of symbols used in
this paper. The relation between the excavation depth
to the maximum wall deflection ratio max /Ze and the
maximum deformation ratio Smax /max are shown in
Figure 4.
East and west wall deflection became large due to
the braced excavation in the middle of Amc layer till
the 4th step, and Smax /max was tend to decrease. However, the wall deflection was decreased gradually, and
Smax /max was tend to increase adversely.
407
Figure 3. Definition of each symbol.
Figure 5. Relation between Ze /B, Smax /max and As /A .
Figure 4. Relation between max /Ze and Smax /max .
The relation between the excavation depth to width

ratio Ze /B, the maximum deformation ratio Smax /max
and the deformation area ratio are shown in Figure 5.
Ze /B is the non-dimensional value for excavation size
and increase due to the braced excavation.
As the increase ratio of the wall deflection and
surface settlement is small till the 4th step due to
the braced excavation in the middle of Amc layer,
Smax /max was tend to decrease. However, Smax /max
was increased conversely due to the braced excavation
after the 4th step. This tendency occurred in the same
way with the relation of As /A .
Figure 6 shows the relation between the wall deflection area A and the backside ground surface settlement area As at the beginning of the excavation and
after removing all struts. More or less A and As
tend to increase from the start to the end. Both side
wall deflections have As
= 0.2 A relationship in the
8th step, and shifted to As
= 0.3 A relationship due
to removal of struts. Unalterably, the west side wall
deflection was larger than the east one.
Figure 6. Relation between A and As .
The relation between the distance from the wall to

backside, wall deflection area A and backside ground
deformation A1 & A2 are shown in Figure 7. Broken
line is the approximate curve based on the calculation
result by least-square method with three measurement
data in each excavation stage.
The ground deformation area attenuated gradually
as getting away from the wall position till the 6th step.
On the other hand, as the increase of the wall deflection
became small, each deformation area changed little.
The attenuation rate of the backside ground deformation area on wall deflection area was just about steady,
about 70% at the No. 1 and about 25% at the No. 2.
408
Figure 7. Influence propagation aspect to backside ground

due to braced excavation.
STATISTICAL COMPARISON BETWEEN

WALL DEFLECTION AND BACK GROUND
SURFACE SETTLEMENT
Figures 8 and 9 show the relation between the wall

deflection area A and backside ground surface settlement area As at the beginning of the excavation and
after removing all struts of all monitoring site for excavation soil characteristic. The broken line shows the
observed data from final excavation stage to all struts
removal stage.
The case of soft and sensitive clayey layer is shown
in Figure 8. For example there were different tendencies between east and west wall deflection behavior,
however, As
= (0.20.3) A relationship was given.
Additionally, the wall deformation and surface settlement increased due to the removal of struts. A and
As were larger than one in the other different grounds.
The case of alluvial sandy layer is struts. A and As
were larger than one in the other different grounds.
The case of alluvial sandy layer is shown in Figure 9.
There are great variances in this reference, however,
Figure 8. Relation between A and As in the case of soft and sensitive clayey layer.
409
Figure 9. Relation between A and As in the case of alluvial sandy layer.
As
= (0.10.4) A relationship was given. The deformation area of wall and surface increased due to
removal of struts the same as the behavior of the soft
and sensitive clayey layer case. But the amount of
deformation was relatively small.
Figure 10 shows the predictive method of backside ground surface settlement based on the relation
between A and As due to the braced excavation.
A = As means that the influence of the wall deflection passes to the surface settlement directly without
volume change (volume loss = 0%).
In the case of including the consolidation settlement, As was larger than A according to the literature.
On the other hand, As was smaller than A in other
cases (JSCE, 1993). Some observed data in the case
of alluvial soft clay ground were added to past examples shown as Figure 10. Because the seepage control
method was adopted in No. 8 Line construction, and
actually the change of ground water level was small,
it is possible to estimate the small amount of degradation for the consolidation settlement caused by
the groundwater level. However past observed data
were distributed along the A = As relation, some
monitoring data in No. 8 Line construction were

located along the As
= (0.20.3) A relation and
under the As
= 0.4 A relation. The backside ground
settlement of this observed data converged smaller
than the past data and the volume loss from A to As
was 7080%. One of the reasons for this phenomenon
is thought to be the support to the neighboring ground
based on the arching effect developed in backside
ground.
5
CONCLUSIONS
The conclusions are drawn as follows:

1. It was found that the earth retaining wall deflection
caused the back side ground deformation like the
circular slip and spread influences to the ground
surface in A-site.
2. The distribution of backside ground surface settlement occurred as convex downward with the
maximum value at some distance from the wall. It
was thought that the friction between the wall and
around ground supported the surface settlement.
410
4. The relation between the wall deflection area A

and backside ground settlement area As was that
As
= (0.20.3) A in prominent soft and sensitive
clayey layer, and As
= (0.10.4) A in prominent
Alluvial sandy layer depended on the construction
condition.
5. Some monitoring data in No. 8 Line construction
were located under the As
= 0.4 A relation. The
backside ground settlement of this observed data
converged smaller than past data, and the volume
loss from A to As was 7080%. One of the reasons
for this phenomenon is thought to be the support to
the neighboring ground based on the arching effect
developed in backside ground.
It is necessary to carry out the evaluation preliminary in detail and use the observational method to
void some kinds of risk. It is believed that this study
will become useful for predicting the influences of
neighboring construction in future.
REFERENCES
Figure 10. Relation between A and As .
These ground movements happened in the 4560

degree area from the bottom of the wall.
3. The attenuation rate of the backside ground deformation area on wall deflection area was just about
steady, about 70% at 2 m distance and about 25%
at 9 m distance from the wall.
Ito, H., Yanagawa, T., Konda, T. & Hayakawa, K. 2006.

Relation between braced wall deflection and deformation
behavior of backside ground due to braced excavation,
Proc. of Tunnel Engineering, JSCE, vol. 16, 439446 (in
Japanese).
JSCE. 1993. Design Calculation Example of Makeshift
Structure based on the Japanese Standard for Cut and
Cover Tunneling, Tunnel Library 4, 107 (in Japanese).
Unei, N., Sakaguchi, Y. & Yamaguchi, H. 2002. Challenge
to Close Approach to or Crossing with Shinkansen Line
and Other Existing Important Structures Construction
plan for subway station on Osaka Municipal Line No. 8,
Tunnels and Underground, Vol. 33, No. 12, 3140 (in
Japanese).
411
Research on the effect of buried channels to the differential settlement of

building
D.P. Liu, R. Wang & G.B. Liu
ABSTRACT: Great and differential settlement occurs under the building during an adjacent metro deep foundation pit construction. Factors influencing the building settlement in different construction stages are analyzed
based on the monitoring data and geology distribution condition. The result shows that the large lateral deformation of retaining wall and time-depended deformation of soft clay under the building are the reasons that make
the major settlement occur. The main factor inducing differential settlement of the building is local distribution of
adverse geological condition. Some feasible means of improvements in practice are presented, which can reduce
effectively the influence of adverse geological condition such as buried channel on excavation and surroundings.
Though the paper the author hopes that engineers can attach more importance to the negative influence of adverse
geological condition, and the results could be valuable reference to other engineering.
Shanghai has the representative soft ground in China.

The stratum of Shanghai mainly made up of saturated
clay, silty clay and sand. Most metro stations and tunnels of Shanghai lie about 20 m beneath the ground surface. Soil of such depth mostly is soft clay which characteristics such as water content, degree of sensitivity,
compressibility and rheology are notable, and the unit
mass, strength and permeability are bad. Except those,
the adverse geological factors such as buried channel, shallow-buried methane, underground barrier etc.
may be exist which will do harm to the constructing of
deep excavation and adjacent buildings. So the adverse
geological factors should be paid enough attention.
During the construction of a metro excavation in
Shanghai, great and differential settlement occurs on
an adjacent building. Factors influencing the building settlement in different construction stages are
analyzed based on the monitoring data and geological distribution condition. Reasons which effect the
major settlement are discussed and the results could
be valuable reference to other engineering.
2
GENERAL SITUATION
The metro station of this study is designed as a two

layers underground frame structure. The total length
of the station is 364.7 m and the width of its standard
Figure 1. Strut profile of standard segment.
segment is 24.5 m.The buried depth of the bottom plate

is 17 m, and the thickness of retaining wall is 800 mm.
As for the edge wells of the station ,the buried depth
is about 18.6 m, the thickness of its retaining wall is
800 mm, Pre-stressed steel pipes of 609 mm in diameter (external) and 16 mm in thickness were installed
at each levels (standard segment: level 1 to level 5;
edge wells : level 1 to level 6) to support the retaining
wall. The protective grade of the excavation is grade
1. The cross section of the standard segment is shown
in Figure 1. Soil profile and geological description of
the soils under the adjacent building of the excavation
are given in Table 1.
413
Table 1.
Profile and geological description of the soils.

Shear strength (peak)
Serial
number
Name
1
1
1
11
12
22
2
filled soil
clay
silty clay
silty clay
clay
silty clay
sandy silt
silty sand
Thickness/m
Bottom
level /m
Water
content /%
Unit
weight /kNm3
Void
ratio
C / KPa
1.2
2.6
4.2
8.0
3.0
6.5
14.5
No penetrated
2.93
0.33
3.87
11.87
14.87
21.37
35.87
No penetrated
34.6
43.0
49.1
38.9
34.9
32.2
26.3
18.2
17.3
16.8
17.6
17.9
18.0
18.8
0.99
1.21
1.39
1.12
1.02
0.94
0.77
21
13
14
16
15
4
1
17.5
17.0
11.0
14.0
18.5
29.0
32.0
Figure 2. Plan of monitoring points and buried channel.
Figure 3. Building settlement with time.
2.1 Adjacent buildings and surrounding condition

The buildings around the excavation are relatively
dense. The prime protection object is the nearer buildings on the south side of the excavation. Great settlement occurred under a teaching building belonged to
a middle school which is sited 14 16 m away from
the excavation on the south side. Total length of the
building parallel to the excavation is 87.882 m and its
width is 15.188 m on the east, 20.883 m on the west.
The building was constructed at the end of 1970s, and
it is the reinforced masonry structure with ring beam,
shallow strip foundation.
According to the geological documents, there are
two buried channels under the building (Figure 2).
Influenced by the adverse geological factor, there has
been about 400 mm total settlement happened before
the excavation construct.
2.2
Building settlement
The curves of building settlement developed with time

are shown in Figure 3 after the excavation constructed
August, 2005. Figure 4 shows the accumulative total
Figure 4. Accumulative total settlement of building during

different stage.
building settlement of several main monitoring points

during different construction stages.
From Figure 4, monitoring points of F13 and
F23 have the large accumulative and differential
414
Figure 5. Building accumulative total settlement in different

stages.
settlement. Until March, 17, the differential settlement in longitudinal direction between F13 and F23
has reached to 45 mm and inclining slope is 0.0005;
and which is 60.7 mm and 0.0041 respectively transversely. Inclining slope of the two directions have
exceeded 0.003 which is the allowable value according
to national criterion. Many cracks appeared in walls
of the building, some local width of cracks is about
10 mm. so did the parallel crakes on the ground surface
between the building and the excavation.
Figure 6. Building settlement velocity in different stages.
BUILDING SETTLEMENT ANALYSIS IN

DIFFERENT STAGES
Great and obvious differential settlement of the building is the result affected by several factors, which have
dissimilar effect in different construction stages. Figure 5 and 6 respectively shows the accumulative total
settlement and settlement velocity with different parts
of the building during the main construction stages.
Based on the main construction stages of the part
of excavation near the building, settlement of different
parts of the building in different construction stages
was analyzed.
3.1
Stage of retaining wall construction
Generally, deformation of the surrounding induced by

retaining wall construction is relatively small and the
effected area is limited. The building lies outside the
influencing area of the retaining wall construction so
that the effect can be neglected when the building
settlement is analyzed.
Field data of the building in retaining wall construction stage is analyzed. Construction of the part of
retaining wall began on Oct.4, 2005 and finished on
Oct.17. During the period building subsided little and
the average settlement is 2.7 mm. Settlement of F21 to

F23 are less than 1 mm and that of F15 to F16 are 1.2 to
1.6 mm, as shown in Figure 4. Take the normal fluctuation of monitoring data into account, the construction
of retaining wall has little influence on building settlement, which tests the conclusion of other correlative
studies.
In Figure 5, settlement in different location of the
building is of variation from the minimum 0.5 mm of
F23 to the maximum 5 mm of F19. Because of the
existing of the buried channel, building settlement of
the location under which is the buried channel are
greater than that of normal subgrade soil. At that time,
there is no other construction, so the main reason leading to the settlement can be ensured as the existing of
the high rheology of the soft soil.
3.2 Stage of excavation
Sequence of the excavation is from the east to the
west. Differential settlement between the two ends was
45 mm till March, 17. In Figure 7, the lateral deflection
of the retaining wall at Q17 is gathered according to
each stage of the excavation. So do the deformation of
excavation and the surroundings as shown in Table 2.
415
In Figure 7, large deformation produced at the early

stage of the excavation. There has been 32 mm lateral
deformation after the 3rd level strut installed which
is mainly related to high time-depended deformation
of the 3rd level soft soil. In addition, long exposure
time without struts and long laying up time after the
installing of struts also made for the lateral deflection
of the retaining wall increasing.
Cooperating with the excavation, basal heave and
the ground surface settlement caused by lateral deflection of the wall are increasing, which also accelerate
the settlement of the building. (Figure 5 and Figure 6)
Settlement velocities of points F6 F8 in the location of buried channel are double greater than those
3.3 Stage of structural construction

Compared with the excavation stage, lateral deflection
of the retaining wall, basal heave and building settlement in this stage are steady relatively because of the
strengthening of the structural stiff. But the building
and ground surface surround the excavation still sank a
little at the rate 0.3 0.4 mm/d. This can be explained
Figure 8. Comparison of settlement velocity between the

parts in buried channel and the others.
Figure 7. Lateral deflection of the wall at Q17.

Table 2.
of points F1 F4. Settlement velocities of points

F13 F15, F17 F20 are also greater than the others. (Figure 8) The maximum local inclining slope of
the building in longitudinal direction is 0.0006, and
which is 0.0052 in transverse. They all exceeded the
allowable value 0.003. So the reason of differential settlement of the building could be mainly attributed to
the local distribution of buried channel.
Deformation of excavation and the surroundings.

Accumulative total on different
moment / mm
Item
Lateral
deflection
of the wall
Basal
heave
Ground
surface
settlement
Q13
Q17
Q19
L6
L7
L8
L9
L10
D65
D66
D67
D68
D69
Stage of excavation
Stage of structure
construction
Increment/
mm
Before the
excavation
The excavation
finished
Structure
construction
finished
Increment /
mm
Velocity /
(mm/d)
31.8
32.7
49.7
16.5
7.7
7.7
24.4
32.1
28.2
35.3
49.6
52.3
70.6
59.1
47.2
61.6
45.9
34.2
49.6
63.0
50.4
52.3
76.9
93.6
86.3
82.2
76.4
60.3
68.6
51.1
35.2
41.0
63.3
60.3
83.2
118.8
130.5
128
130.0
27.3
14.5
11.9
29.4
26.5
41.8
38.6
18.3
24.1
41.6
44.0
34.0
11.6
1.05
0.60
0.70
1.13
1.02
1.67
2.27
1.66
0.93
1.60
1.76
2.0
1.05
416
17.9
13.1
7.0
5.2
1.0
8.6
0.3
9.9
30.9
41.9
36.9
41.7
47.8
Velocity /
( mm/d)
0.15
0.13
0.07
0.18
0.03
0.28
0.01
0.23
0.30
0.41
0.35
0.37
0.41
as the reason of time-depended deformation of the soil

around the excavation.
REASONS ANALYSIS
1. Relatively great lateral deflection of the retaining

wall causes the whole settlement of ground surface.
Time-depended deformation of the soft soil disturbed from the construction, overload of the building, long exposure time without struts and long
laying up time after installing the struts made the
lateral deflection of the retaining wall increasing.
2. Time-depended deformation of the soft soil
under the building increased the whole building
settlement.
Soft soil has obvious rheology and sensitivity.
Time-depended curve of typical soft soil in Shanghai is shown in Figure 9. High time-depended
deformation because of the disturbing of excavation construction causes the sustaining settlement
of the building.
3. Differential settlement of the building mainly
attributes to the local distribution of adverse geological action.
Settlements in different location of the building are
obviously of variation because of the local distribution of buried channel, the building settlements
of the part under which is the buried channel are
greater than those located on normal subgrade soil.
CONCLUSION
1. Optimizing the design of the retaining system of the

excavation in complex geological condition, shortening the exposure time of retaining wall during
excavation should be done to control the lateral
deflection of retaining wall, which in turn could
reduce the effect on the surroundings.
2. The adverse geological condition such as buried

channel should be paid enough attention. The
buried channel should be strengthened or separated
from the building to decreasing the disadvantageous effect on the building in practice.
3. Monitoring data can give feedback to the designers about the construction so that they can adjust
design parameters in time to instruct the construction and ensure the safety of the excavation and the
surroundings.
REFERENCES
Gao, Y.B., Wu, X.F. & Ye, G.B. 2003. Study on the Influence
of Diaphragm Wall Construction on adjacent Building
Settlement, Underground Space. 23(2):115118.
Liu, G.B. & Lu, H.X. 2004. Study on the Influence
of Building Settlement Diaphragm Wall Trench Construction, Chinese Journal of Geotechnical Engineering.
26(2):287289.
Liu, J.H. & Hou, X.Y. 1997. Manual of Excavation engineering, Chinese Construction Industry Press, Beijing.
Tang, H.F. & Li, B. 2004. Analysis of measured environmental influence and protection of specially big and deep
excavation engineering in soft soil areas, Rock and Soil
Mechanics. 25(s1):553558.
Xiong, J.H. & Lou, X.M. 2004. Monitoring and Analysis of displacement of a Foundation Pit in Soft Soil,
Underground Space. 24(3):354358.
Zhang, L 2004. Analysis of the Settlement of Buildings
Caused by the Construction of Metro Pit under Special Geological Condition, Urban Railway Traffic Study.
23(5):3841.
Zhu, R.J. , Gao, Q & Qi, G. 2006. Settlement Analysis/n
Building near Reining Pile of Deep Foundation Pit,
Journal of Chongqing Jianzhu University. 28(2):5255.
Figure 9. Dynamic triaxial rheology test of soft clay.
417
Performance of a deep excavation in soft clay

G.B. Liu & J. Jiang
Department of Geotechnical Engineering, Tongji University, Shanghai, P.R. China.
C.W.W. Ng
Department of Civil Engineering, the Hong Kong University of Science and Technology, HKSAR
ABSTRACT: The observed response of diaphragm wall and the surface settlement of a deep excavation for a
metro station in Shanghai soft clay are presented, compared with the similar excavations (retained by a diaphragm
wall) in soft clay areas in Asia and five metro station excavations (similar depth) in Shanghai. Results show that
the maximum lateral wall deflection (hm ) and ground settlement (vm ) are 0.32% and 0.1% of final excavation
depth (He ), respectively. Lateral wall deflections are near the average magnitude value in Shanghai, smaller than
those in Singapore. The largest lateral wall deflection is near the excavation center and the ratio of hmcor /hmcen of
the maximum wall deflection is 0.390.74. The ground settlement is relatively small and falls in the Zone I limit
described by Peck (1969). The adoption of prestressed steel struts, short excavation sections, fast workmanship
sequences and compaction grouting may contribute to the deformation characteristics of deep excavations for
metro stations in Shanghai.
Excavations for metro stations are in general deep,

long and narrow, and located in urban environments. In
Shanghai, China, some metro lines have been built and
several lines are under construction or planned, to cross
the large and congested city. Thick soft-to-medium
alluvial and marine deposit strata exist in this area.
Under such circumstances, movement control, rather
than stability at the deep excavation site becomes the
governing factor for design and construction (Liu &
Hou 1997).
Research shows that field monitoring and performance analysis are essential to deep excavations in
soft soils. Monitoring the field performance of the
excavation and the magnitude of movements in the
surrounding soil (Peck 1969; ORourke 1976; Clough
et al.1989) along with estimating the effects of such
movements and field performance on adjacent structures (Burland & Wroth 1974, Boscardin & Cording
1979) are important to design engineers. The observed
performance of deep excavations and case histories are
very useful to verify design assumptions and reduce
the construction risk in the excavation process (Whittle
et al. 1993; Ng 1998, Ou 1998, Long 2001, Finno &
Bryson 2005, 2007, Leung & Ng 2007).
The observed responses of diaphragm walls and the
ground surface of a deep multistrutted metro station
excavation in Shanghai soft clay are presented here.

Diaphragm wall deformations and ground movements
are compared with some case histories in Shanghai and
excavations in soft clay in other locations inAsia. From
the measured data in this study, some recognition of the
characteristics of deep excavations can be obtained and
can offer references for back analysis and case studies.
SITE CONDITION
2.1 Introduction
Shanghai is located in the front fringe of the Yangtze
River Delta in China. The Da Muqiao excavation is
situated in the southwest of Shanghai and there is an
interchange for two lines, Metro 4 and Metro 9. The
site plan is shown in Figure 1. Note that the excavation
width is variable.
The site crosses two roads, Da Muqiao Road and
Ling Ling Road. The shadowed area in Figure 1 shows
the interchange excavation (35 m 25 m 20.7 m),
in the shape of cross. The excavation depth is
3.8 m deeper than the excavation depth in other parts
(16.9 m). There are several concrete, three-story to
six-story buildings along the sides of the excavation.
Figure 2 shows a cross section of section 1-1 of the
interchange part (INP); including the support system,
419
Figure 1. Site plan of the metro station and excavation direction.
Figure 3. A typical soil profile and soil parameters.
Figure 2. Cross section of the INP (Section 1-1).
the strut level and excavation depth. The concrete

diaphragm wall for the INP is 40 m deep and 0.8 m
thick; the 16.9 m deep excavation is retained by a 34 m
deep and 0.8 m thick perimeter concrete diaphragm
wall. The ground level of the site is 3.75 m above the
Shanghai City Datum (SCD). There is 3 meter compaction grouting below the final excavation to increase
that soil strength in this part. Six steel struts were used
at the INP in vertical plane while there were five steel
strut levels in other excavation parts. All struts are steel
tubes and the first one has a 580 mm diameter and other
ones are 609 mm in diameter and are 16 mm thick.
Two steeltubes (12.45 m) were inter-joined to form the
struts (24.6 m). The struts have a 3 m lateral spacing.
The specific installment of struts was described by
Wang (2005).
2.2
Geology
The strata of Shanghai are thick soft soils comprising

Quaternary alluvial and marine deposits. A high water
content, low shear strength, high compressibility and

low ground bearing capacity are typical characteristics
of the soft soil in Shanghai.
The ground is underlain by relative soft to medium
soft marine deposits, which generally consist of uniform bedding planes. A typical soil profile from up
to down obtained during site investigation is shown
in Figure 3. The top layer is artificial fill, 1.6 m in
thickness. About 45 m thick clay underlies the fill. The
inside of the excavation is among the clay, which is
a quaternary alluvial and marine deposit. In general,
groundwater conditions are approximately hydrostatic
1.0 m below ground level.
CONSTRUCTION PROCEDURE OF THE

EXCAVATION
From installing the diaphragm wall to the completion of casing the bottom slab, the working period
is divided into eight main construction stages, listed
in Table 1. The excavation depth of the INP is 3.8 m
deeper than other parts (see Figure 2) so there is one
more excavation stage indicated by italics Table 1.
420
Summary of the main construction.
Stage
(1)
Construction operation
(2)
Day
(3)
1
2
3
Construction of diaphragm wall

Construction of bored pile
Reduce level dig to L1 & installation
of L1 strut
Excavation to L2 & installation of L2 strut
Excavation to L6 & casting bottom slab
Excavation to L7 & casting the bottom slab
81
116
122
4
5
6
7
8
9
Excavation depth (m)
Table 1.
129
136
143
151
160
168
The excavation of this long metro station was

divided into several short sections in the horizon plane.
Note that the excavation was from the two short sides
to the middle part. The INP and the east part were synchronously excavated, indicated as Part 1. Part 2 was
at the west end and was excavated one month after
Part 1 was started. After finishing these two excavations, Part 3 was excavated from both ends. A short
section of soil excavation was adopted in Part 3 also.
Each excavation depth was dug to 0.3 m below the strut
level to offer space for installing the struts.
4
INSTRUMENTATION
Inclinometer tubes were fixed to the steel reinforcement cages and concreted in the retaining walls. The
casing was installed with a pair of grooves oriented in
the expected direction of movement. To avoid twisting
of the casing, the grooves were checked before installment to ensure that their twist angle was no more than
0.1%. The casing was fixed tightly to the steel cages of
the retaining wall. The groove direction was checked
again after lowering the cages into the bentonite. The
twist angles of the grooves were measured after the
concreting. The allowable value of each twist angle
was 0.2%. Probes with a resolution of 0.02 mm/50 mm
and temperature of 20 to +50 were used in the
inclination monitoring.
Ground surface settlement monitoring markers
were located on the sides of the excavation. The markers were located about 1.535 m perpendicularly away
from wall at each rectangular box station and were
secured 0.5 m below the ground surface. The section
settlements were monitored with a leveling instrument
with a stated accuracy of 0.5 mm/km.
5
5.1
Lateral wall deflection (mm)

20
40
60
10
20
30
I1
I3
I5
I7
I9
I2
I4
I6
I8
40
Figure 4. Maximum lateral wall deflection at the final
excavation depth.
the lateral wall deflection. Only nine of these inclinometers are pictured, as I1 to I9, in Figure 1, representing three typical locations of the diaphragm walls.
(1) Inclinometers I1, I2, I5, I6 and I7 were at or near the
corners; (2) Inclinometers I3 and I4 were at the centers of the excavation; (3) Inclinometers I8 and I9 were
set for the INP. These nine inclinometers measured the
data that are analyzed here for lateral wall deflection.
Figure 4 shows the lateral wall deflection at I1 to
I9 at the end of the final excavation (Stage 8). The
upper final excavation depth is that of the general part
while the lower one is that of the INP. As indicated
in this figure, the profiles of lateral wall deflection
at the inclinometers are similar but with different
deformation magnitudes.
The lateral wall deflection of type (1) is smaller than
type (2). This finding shows that wall deflections were
affected by the stiffness of the corners. The maximum
wall deflection of this deep excavation was at the center
of excavation I3, with a value of 54.5 mm; while the
minimum wall deflection was at I5, with a value of
20.2 mm. I5 is at Corner 3 and the excavation width
on the west side (see Figure 1) is significantly shorter
than that on east side. This varied excavation width
may contribute to the smallest wall deflection at I5.
The lateral wall deflection of type (3) (INP) is
between type (1) and type (2). The maximum values
for I8 and I9 were 38.2 mm and 51.0 mm, respectively.
Note that the locations of all lateral wall deflections
were above the final excavation level.
5.2 Relationship between maximum lateral wall
deflection and excavation depth
OBSERVED WALL DEFLECTION

Lateral diaphragm wall deflection
Along the south side of this deep and long excavation, seventeen inclinometers were placed to monitor
Lectures on lateral retaining wall deflection caused by

deep excavations in Asian soft clay were reported by
some researchers (Ou et al. 1993, 1998, Wong et al.
1996, Tamano et al. 1996, Lee et al. 1998, Wang et al.
421
Summary of the six metro excavations.

Strut
No.
He
(m)
Name
Pudian
Yangshupu
Pudongdao
Luban
Pudong
Damuqiao
16.3
16.5
16.5
18.2
17.3
16.9
4
4
4
5
5
5
Ks
626.4
626.4
626.4
777.6
1101.6
1414.3
FOSbase
2.58
2.59
2.61
2.35
2.53
2.43
0.6
hm /
He (%)
0.5
0.42
0.57
0.15
0.40
0.28
0.32
0.4
0.2
0.1
Maximum wall deflection (cm)
(2.53)
(2.43)
(2.61)
FOSbase=3.0
This study
Design curves (Clough et al. 1989)
100
1000
10000
Ks
18
hm/H=0.9%
15
12
hm/H=0.6%
6
3
0
(2.35)
FOSbase=2.0
0.3
0.0
10
(2.59)
FOSbase=1.4
(2.58)
hm/He (%)
Table 2.
UOB, Singapore
Bugis, Singapore
Singapore Parking
CTE, Singapore
Newton, Singapore
Formosa, Taiwan
Taiwan Chi Ching
TNEC, Taipei
Osaka A, Japan
5 cases inShanghai
This study
Figure 6. Normalized lateral wall deflection (hm /He ) versus system stiffness (Ks ).
5.3 Effect of the stiffness of the support system on

the maximum wall deflection
The effect of the stiffness (KS ) of the support system on
lateral wall movement was proposed by Clough et al.
(1989). KS is expressed as follows:
hm/H=0.14%
hm/H=0.3%
10
20
30
Excavation depth (m)
40
Figure 5. Relationship of the maximum lateral wall deflection to the excavation depth.
2005, Wallace, 1992, Hulme et al. 1989,) Figure 5

shows the measured lateral wall deflection compared
to the excavation depth of this deep excavation. Data
of five other deep excavations in Shanghai (listed in
Table 2.) and several excavations in Taiwan, Japan and
Singapore soft clay are plotted in Figure 5 for comparison. Note that all the excavations were in soft clay
and the retaining walls were all diaphragm walls.
Figure 5 shows that the magnitude of the maximum
wall deflection in Shanghai soft clay is smaller than
that in Singapore soft clay while it is similar to that
in Taiwan and Japan soft clay. The ratio of hm /He
in metro excavations in Shanghai is between 0.14%
0.6%, while that in Singapore is between 0.3%0.9%,
and that in Taipei and Japan is in 0.3%0.6%. The
ratio of hm /He in this study is 0.32%. A comparison of geotechnical parameters of the soil in Shanghai
clay and Taiwan clay shows that undrained strength
and water content are similar, but the average measured deflections in the Shanghai soil were smaller
than in the Taiwan soft soil. The use of the multistrutted support system, the prestressed steel struts,
the short excavation sections and the fast workmanship sequences in Shanghai probably account for these
differences.
where EW is the Youngs modulus of the wall; I is

the second moment of the area of the wall section,
I = t 3 /12, t is the wall thickness; h is the average vertical prop spacing of the multistrutted support system
and W is the unit weight of water.
Figure 6 shows the normalized lateral wall deflection of the long excavation in the chart as suggested by
Clough et al. (1990). Another five metro excavations
with the same retaining wall type and support system in Shanghai with excavation depths from 16.3 m
to 18.2 m, are chosen for comparison. A summary of
these metro excavations is listed in Table 2. The retaining structures of all these excavations were 0.8 m width
diaphragm walls and the value of EW was 30 kN/mm2 .
This form by Clough (1990) is for excavation in softto-medium clay to account for system stiffness and
FOSbase . The value of FOSbase is calculated by the
definition proposed by Clough et al. (1989).
The Ks value ranges from 626.4 to 1414.3 and the
ratio of hm /He varies from 0.15 to 0.57 with the value
of FOSbase from 2.35 to 2.61. However, for a given
Ks , there is a relatively scatter in hm /He values, there
may be no very strong correlation between measured
hm /He and Ks value. It can be seen that measured lateral wall deflection decreases with increasing FOSbase .
With similar system stiffness, the larger value of
FOS base the smaller the ratio of hm /He . And the measured data do not seem to correspond very well with
the design curves. The suggested chart only could give
422
Distance from east corner (m)

110
220
Normalized distance from wall to

excavation depth
330
0
0.0
I5
20
30
I1
I2
I7
I8
40
I4
I8
I6
50
I9
60
Normalized settlement to
excavation depth (%)
hm (mm)
10
I3
0.2
0.4
1.0
1.2
1.5
0.05
0.6
0.8
Peck (1969)
0.1
TEN, Taipei
IMM Building, Singapore
Fareast, Taipei
Settlement in Shanghai
Figure 7. Maximum lateral wall deflection versus.
preliminary predictions of wall movement in six metro

excavations in Shanghai.
5.4 Three-dimensional deformation of excavation
Researches on three-dimensional responses of excavations by measured data or finite analysis (Bono
et al.1992, Ou et al. 1993, Wong 1993, 1996, Lee
et al. 1998, Finno et al. 2007) showed that ratios of
corner /center in deeper excavations are smaller in general. Figure 7 shows the measured maximum lateral
wall deflection (hm ) at the end of the final excavation in this study. The lateral wall deflection at and
near the center of the excavation is larger than the wall
deflection near or at the corners, which means that
the stiffening effect of the corners affected the deformation of the diaphragm walls. Note that the reduced
lateral wall deflection at I5 is significant. The ratio of
each maximum wall deflection to the maximum wall
defection at the center, hmcor /hmcen , of the wall was
0.390.74.
6
SURFACE SETTLEMENTS
In some cases, settlement was measured after the

starting of construction of the diaphragm wall; some
were measured at different times. Making consistent
comparisons is therefore difficult. In this study, the
settlement of the ground surface during and after the
final excavation was measured.
Eight groups of ground settlements were measured
along one of the long sides (south side) and two ends
of the excavation (see Figure 1). All data were selected
from one side of the site. Figure 8 shows the relation
of the maximum surface settlements to the distance
from the diaphragm walls at the final excavation depth
of this study.
In comparison with surface settlements in Singapore and Taiwan soft clay, the measured data in this
deep excavation fell into a relatively small range,
despite the weak ground conditions in Shanghai.
Figure 8. Distribution of the ground surface settlement

normalized by the excavation depth.
Wang (2005) summarized the surface settlements of

some metro excavations in Shanghai and also indicated that the magnitude was relative small (hm /He is
less than 0.1%), plotted in Figure 8 also.
The 3 m thick compaction grouting below the final
excavation level, which improved the ground condition, and the high-stiffness diaphragm wall and
prestressed multistrutted steel struts and fine workmanship may contribute the small surface settlements.
7
DISCUSSION OF THE MOVEMENT

MECHANISM
Compared with the lateral wall deflection and ground

settlement caused by excavation in soft clay in Singapore, Japan and Taiwan, the measured deformations
in this deep excavation for a metro station in Shanghai is relatively smaller (see Figure 7). The 3 m
thick compaction grouting below the final excavation
level, usage of prestressed multi struts and a stiffing
diaphragm wall, the short soil excavation sections and
quick and fine construction are main factors associated
with this measured result. Among the above reasons,
one that ought to be noted is that the vertical spacing of
the struts in the Shanghai excavation was smaller than
that in Singapore and Taiwan, which contributed to the
small wall deflection. Generally in 12 m to 18 m deep
Shanghai metro excavations, four to five struts are set
to support the system, while in the 9.9 m deep excavation of CTE in Singapore, for example, only two struts
were set in the vertical plane. The system stiffness in
the Shanghai excavation was larger; it is not surprising
to observe the smaller deformation in Shanghai.
CONCLUSIONS
Observed data from a deep and long metro excavation, with an interchange metro station (35 25 m2 )
423
crossing it, in Shanghai soft clay is presented. On

the basis of the interpreted observed field data, the
following conclusions are drawn:
1. The wall deflections along the long side and three
corners are different, but with similar profiles. The
ratio of hm /He in this deep excavation is less than
0.32%. The measured maximum wall deflection
is near the center. Three-dimensional analysis of
this deep excavation shows that the corner effect is
found. The hmcor /hmcen of the wall was 0.390.74,
though the aspect ratio was large.
2. Compared with the excavations in soft clay in Taiwan, Japan and Singapore retained by diaphragm
walls and another five similar metro excavations
in Shanghai, this maximum lateral wall deflection
falls near the average magnitude for Shanghai and is
at the low limit line for Singapore. The undrained
strength and water content of Shanghai clay and
Taiwan clay is similar, but the average measured
deflections in Shanghai are smaller than in Taiwan.
The use of prestressed steel struts, a support system, short excavation sections, a fast workmanship
sequence, and compaction grouting below the final
excavation level in Shanghai may contribute to this
smaller deflection.
3. The influence of support stiffness on wall movement was analyzed.A smaller lateral wall deflection
of excavation with large Ks values was observed.
Though scatter exists in the chart proposed by
Clough et al. (1990), the suggested chart seems to
allow preliminary predictions of wall movements
in excavations for metro stations in Shanghai soft
soil.
4. The settlement along the corners and the long side
is relatively smaller (hm /He is less than 0.1%) than
that in other Asian soft clay, though the soft clay
strata is thick in this deep excavation. This settlement is located in the zone I as proposed by Peck
(1969), which is consistent with the case study in
Shanghai (Wang et al. 2005).
ACKNOWLEDGMENTS
The authors would like to thank colleagues who contributed to the field monitoring in Shanghai and
acknowledge the earmarked research grant 618006
provided by the Research Grants Council of the
HKSAR.
REFERENCES
Bono, N.A., Liu, T.K. & Soydemir, C. 1992. Performance of
an internally braced slurry-diaphragm wall for excavation
support. Slurry walls: Design, construction, and quality
control, ASTM STP 1129, D.B.Paul, R.G.Davidson, and
N. J. Cavalli, eds., ASTM, Philadelphia. pp.169190.
Boscardin, M.D. & Cording, E.J. 1979. Case studies of

building behavior in response to adjacent excavation. University of Illinois Rep. for the U.S. Department of Transportation, Rep. No. UMTAIL-06-0043-78-2, Washington
D. C.
Burland, J.B. & Worth, C.P. 1974. Settlement of buildings
and associated damage. Proceeding of conference on
settlement of structures, Cambridge. pp. 611654.
Clough, G.W., Smith, E.M. & Sweeney, B.P. 1989. Movement control of excavation support systems by iterative
design. In proceeding of current principles and Practices
on foundation and engineering, ASCE, New York. Vol. 2,
pp.869884.
Clough, G.W. & ORourke, T.D. 1990. Construction induced
movement of in-situ walls. In proceeding of design and
performance of earth retaining structures. ASCE special
conference, Ithaca, N.Y. pp. 439470.
Finno, R.J., Bryson, L.S. & Calvello, M. 2005. Performance
of a stiff support system in soft clay. Journal of Geotechnical and Geoenvironmental Engineering, ASCE, 128(8).
pp. 660671.
Finno, R.J., Blackbum, J.T. & Roboski, J.F. 2007. Threedimensional effects for supported excavation in clay. Journal of Geotechnical and Geoenviromental Engineering,
ASCE, 133(1). pp. 3036.
Hulme, T.W., Potter, J. & Shirlaw, N. 1989. Singapore MRT
system: Construction. Proc., Instn. of Civ. Engrs., Vol. 86,
London, 709770.
Lee, F.H., Yong, K.Y., Quan, K.C. & Chee, K.T. 1998. Effect
of corners in strutted excavations: field monitoring and
case histories. Journal of Geotechnical and Geoenviromental Engineering, ASCE, 124(4). pp. 339349.
Leung, E.H.Y. & Ng, C.W.W. 2007. Wall and ground movements associated with deep excavations supported by
cast in situ wall in mixed ground conditions. Journal of
133(2). pp. 129143.
Liu, J.H. & Hou, X.Y. 1997. Excavation engineering hand
book. Chinese Construction Industry Press, Beijing, P.R.
China.
Long, M. 2001. Database for retaining wall and ground
movements due to deep excavation. Journal of Geotechnical and Geoenviromental Engineering, ASCE, 127(3).
pp. 203224.
Ng, C.W.W. 1998. Observed performance of multi-propped
excavation in stiff clay. Journal of Geotechnical and
Geoenviromental Engineering, ASCE, 124(9). pp. 889
905.
ORourke, T.D., Cording, E.J. & Boscardin, M.D. 1976. The
ground movements related to braced excavation and their
influence on adjacent structures, Univ. of Illinois Rep. for
the U.S. Dep. of Transportation, Rep. No. DOT-TST-76T22, Washington D.C.
Ou, C.Y., Liao, Hsied. P.G. & Chiou, D.C. 1993. Characteristics of ground surface settlement during excavation.
Canadian Geotechnical Journal, 30(5). pp. 758767.
Ou, C.Y. Liao, Hsied. & Lin, H.D. 1998. Performance of
diaphragm wall constructed using top-down method. Journal of Geotechnical and Geoenviromental Engineering,
ASCE, 124(9). pp. 798808.
ground. In proceeding of the 7th international conference
on soil mechanics and foundation engineering, Mexico
City. Vol. 1, pp. 225281.
424
Tamano,T., Fukui, S., Mizutani, S.,Tsuboi, H. & Hisatake, M.

1996. Earth and water pressures acting on a braced excavation in soft ground. Proc., Int. Symp. Geo Aspects of
Underground Constr. in Soft Ground, City University,
London, 207212.
Wallace, J.C., Ho, C.E. & Long, M.M. 1992. Retaining
wall behavior for a deep basement in Singapore marine
clay. Proc., Int. Conf. Retaining Struct., Thomas Telford,
London, 195204.
Wang, Z.W., Ng, C.W.W. & Liu, G.B. 2005. Characteristics of wall deflections and ground surface settlements
in Shanghai. Canadian Geotechnical Journal, 42(10).
pp. 12431254.
Whittle, A.J., Hashaha,Y.M. & Whitman, R.V. 1993. Analysis
of deep excavation in Boston. Journal of Geotechnical
Engineering, ASCE, 119(1). pp. 6990.
Wong, L.W. & Patron, B.C. 1993. Settlements induced by
deep excavations in Taipei. 11th Southeast Asian Geotechnical Conf., Singapore. pp. 787791.
Wong, L.W., Poh, T.Y. & Chuash, H.L. 1996. Analysis of
case histories from construction of the central expressway in Singapore. Canadian Geotechnical Journal, 33(1).
pp. 732746.
SYMBOLS
EW =Youngs modulus of the wall;
FOSbase = factor of safety against the basal heave;
h = average vertical prop spacing of multi-strutted
support system;
He = final excavation depth;
I = second moment of inertia of the wall section;
INP = interchange part excavation;
Ks = system stiffness;
t = the wall thickness;
W = unit weight of water;
corner = deformation at corner of excavation
center = deformation at center of excavation
hm = magnitude of maximum horizontal diaphragm
wall deflection;
hmcor = maximum lateral wall deflection at corner;
hmcen = maximum lateral wall deflection at center;
vm = maximum ground surface settlement;
425
Deformation monitoring during construction of subway tunnels in

soft ground
S.T. Liu
Department of Surveying and Geo-Informatics, School of Civil Engineering, Tongji University,
Department of Civil Engineering, Henan Institute of Engineering, Zhengzhou, P.R. China
Z.W. Wang
Fourth Team of Henan Province Coal Field Geology Bureau, Xinzheng, P.R. China
ABSTRACT: Monitoring the deformations of subway tunnels in soft ground is a principal means for selecting
the appropriate excavation and support methods in the design. In this paper, briefly describes the basic deformation
monitoring requirements, analyzes the tunnel deformation monitoring accuracy difference between mountain area
and urban area, discusses the measurements of deformations at ground surface, measurements of deformations
on the ground and measurements in the tunnel, especially gives an example of vertical convergence monitoring,
introduces the data processing method and the quality evaluating method based on the measured data. Mainly
explores the technique of tunnels crown settlement monitoring, convergence monitoring and the Automatic
Deformation Monitoring methods. It is put forward that the monitoring job during the construction of subway
tunnels can be done with both traditional and modern surveying instruments and the job will be significant to
the same kinds of projects.
INSTRUCTIONS
Determining shape and position changes occurred

in engineering is one of an application area of
geodetic surveys. Temporary and permanent deformations occur in engineering structures such as dams,
bridges, tunnels, viaducts and towers due to natural
and artificial forces. Causes of these deformations
are usually physical properties of ground, weight of
structure, active external forces etc. (Clough 1960).
The need to upgrade and further develop transportation infrastructure has lead to the on-going construction of large-diameter, long tunnels under difficult
conditions.
Such conditions usually arise from a combination
of adverse ground and groundwater regimes, very high
overburden pressures or, in the case of urban tunnels,
the existence of sensitive structures within the zone
of influence of the tunnel. It is imperative to provide
accurate and frequent monitoring of tunnel linings to
detect any movements that could pose a safety hazard.
Deformation monitoring in tunneling usually
includes some of the following measurements
(Dunnicliff 1993):
1. Convergence of the tunnel wall, and usually crest
settlement and spring-line closure.
2. Deformations at ground surface including settlements and tilts of surface structures.

3. Deformations in the ground around the tunnel.
Tunnel deformations can be monitored with geodetic or geotechnical methods. Geodetic measurements
provide absolute coordinates of the target locations in
time, while geotechnical measurements usually provide relative displacements of the target locations with
respect to an initial condition (at the time when the
initial measurement is recorded) (Hisatake 1999).
Geotechnical measurements can provide absolute
coordinates of the target locations in time, if the
initial positions of the targets are obtained using
geodetic means. Depending on the tunneling application, deformation measurements can be recorded,
processed and evaluated in real-time using digital
recording and telecommunication systems, or can
be recorded manually and processed later in batch
mode.
Real-time processing of ground deformations offers
the possibility of rapid response to upcoming situations but requires advanced technology and an
appreciably higher cost. Thus, real-time monitoring
is limited to cases where rapid response is absolutely
necessary, i.e., mainly in urban tunnels near sensitive
structures.
427
In this paper, briefly describes the deformation

monitoring requirements, and then discusses the deformation monitoring at ground surface, in the ground
and in the tunnel separately, furthermore gives an
example of traditional vertical convergence monitoring, at last mainly introduces the Automatic Deformation Monitoring method.
2
BASIC REQUIREMENTS
The basic requirements are: lay out monitoring benchmarks, monitoring targets timely, take measurements
regularly, analyzes the survey data and feedback information to relative departments in time (Kavvadas
1999).
Measurement accuracy depends on the purposes of
the deformation observation. In order to ensure monitoring accuracy, Operating personnel must be familiar
to the purposes and equipment operating rules, survey
team members should corporate and work carefully.
Each monitoring step should abide the four fixed
principle (Liu 2006). The so-called four fixed principles namely: instruments and rods fixed, observing
staff fixed, monitoring method fixed and monitoring
environment fixed.
The first step in deformation monitoring is to carry
out a basic survey, consisting of individual surveys
inside a narrow time frame. All following surveys will
be compared to the results of this survey. It is therefore important that this survey is carried out before the
deformations to be observed can be expected to occur
and it is also important that the basic survey is carried
out with a high degree of reliability regarding results,
as it is obviously impossible to follow up with checksurveys after deformations have taken place or can be
expected to have taken place. It is recommended that
the first elevation result should be figured out in the
average of at least two times of monitoring (Liu 2007).
The equipment being used should be tested before
the first time monitoring, after continuously use for
36 months the equipments should be tested again
(Kaiser 1993).
Benchmarks and stations and Monitoring points
inside the tunnels should be set in area where rocks
are stable. Benchmarks and stations inside the tunnel
are installed in special brackets, which forced centered
(Kontogianni & Stiros 2002, 2003).
3
3.1
MONITORING ACCURACY DIFFERENCE

BETWEEN MOUNTAIN AREA AND
URBAN AREA
Objectives difference
The objectives of ground deformation monitoring are

different in mountain and urban tunnels. In mountain
tunnels, the main objective of deformation measurements during construction is to ensure that ground
pressures are adequately controlled, i.e., there exists an
adequate margin of safety against collapse, including
roof collapse, bottom heave, failure of the excavation
face, yielding of the support system, etc. (Mihalis &
Kavvadas 1999).
Adequate control of ground pressures ensures a safe
and economical structure, well adapted to the inherent
heterogeneity of ground conditions. This procedure is
compatible with modern tunnel design methods which
include a range of excavation and support systems to
cover the anticipated spectrum of conditions along the
tunnel, with selection of the applicable system in each
case relying on the encountered geology at the tunnel face, experience on tunnel behavior at previously
excavated sections under similar conditions and, on
accurate deformation measurements, i.e., by applying
the so-called observational method.
This method of construction can ensure adequate
safety and, at the same time, an economical construction. On the contrary, in urban tunnels, the main
objective of ground deformation monitoring is to limit
ground displacements to values sufficiently low to
prevent damage to structures and utilities at ground
surface. Thus, the fundamental difference in deformation monitoring stems from the fact that in mountain
tunnels the objective is to guard against an ultimate
limit state (i.e., collapse) while in urban tunnels the
objective is to guard against serviceability limit states
(i.e., crack initiation) for structures and utilities at
ground surface.
3.2
Monitoring accuracy difference
As a result of these differences in objectives, design

philosophies, and construction techniques, the types
and required accuracy of the measured ground deformations vary between the two classes of tunnels, as
follows.
1. In mountain tunnels, considerable ground deformations are deliberately permitted (and often provoked) in order to reduce the initially very large
geostatic loads on the temporary support by
increasing ground de-confinement. Such reduction of ground loads on the tunnel support can
be appreciable and, thus, extremely beneficial provided that excessive loosening of the rock mass
is prevented (such loosening can cause roof failures and an eventual increase of the ground loads).
De-confinement is achieved by controlled inward
ground deformation at the excavation face (facetake), controlled delay in the completion of the
temporary support measures (by increasing the distance from the face where the tunnel invert is
closed), a relatively flexible temporary support system (e.g. long passive rock-bolts and thin sprayed
428
concrete liners) and, finally, by installing the permanent lining at a later time when evolution of
the long-term (creep) ground deformations has
practically stopped. In extreme cases of strongly
squeezing ground conditions, sliding supports may
be installed to permit tunnel wall convergences of
several tens of centimeters. In all these cases, control of ground deformations depends strongly on
efficient and timely deformation measurements.
However, due to the large ground deformations
(several centimeters and even several tens of centimeters), the required level of precision of these
measurements needs not be excessive; typically,
accuracy of the order of one centimeter is sufficient
in mountain tunnel applications.
2. In urban tunnels, the main objective is limiting ground deformations around the tunnel and
thus causing the minimum possible movement and
disturbance at ground surface and the structures
founded there. This is achieved by (a) limiting
inward ground deformation at the excavation face
(face-take), e.g. by face pre-reinforcement using
fiber-glass nails, stiff steel beams (fore-poles),
cement- or jet-grouting techniques, (b) by installing
a stiff temporary lining, usually including invert
closure, as early as possible and (c) by installing
the final lining as quickly as possible, especially
when tunnel wall convergences continue to evolve
with time. The above stiff construction methods
tend to reduce ground de-confinement and thus the
ground loads on the tunnel lining are a significant
fraction of the initial geostatic loads are much
smaller than those in deep mountain tunnels. Due to
the small ground deformations induced by tunneling (usually less than 10 mm at ground surface and
occasionally less than 5 mm), measurement precision and the early installation of the measuring
devices is of utmost importance.
is typically 0.2 mm (over about 100 m lengths) for precise leveling and 1 for angles and (1 mm + 2 mm
106 D) for distances in the case of faade monitoring with total stations. Faade monitoring can be
automated and measurements can be obtained and
transmitted in practically real-time.
Inside buildings and in areas with limited visibility
for the application of the above geodetic measurements, geotechnical measurements can be performed
using the following precision instruments (Moaveni
2003):
1. Electronic Liquid Level gauges, for the measurement of settlements at several locations. The
method consists of installing a number of liquid
filled pots, hydraulically connected to a reference
pot located in a stable area. The elevation of the
liquid in the reference pot is maintained constant
by means of a mini-pump, reservoir and an overflow unit. LVDT float sensors monitor the height
of the liquid in each pot. When settlement or heave
occurs, the sensor detects the apparent change in
the height of the liquid and transmits the signal to a
data logger for continuous monitoring and real-time
processing. The accuracy of the system is 0.3 mm.
2. Electrolytic tilt sensors are precision bubble levels
that are electrically sensed as a resistance bridge.
The bridge circuit outputs a voltage that is proportional to the tilt of the sensor. The sensors
are usually attached on metal beams, one to three
meters long, with their ends mounted on the structural elements to be monitored. Chains of such tilt
sensors are often installed in sequence in the horizontal direction to monitor differential settlements
along long walls or beams. The precision of the
instrument is typically1 .
3. Surface clinometers (tilt meters), precise taping
using invar tape between fixed anchor points and
various types of crack-meters are also used in
measuring deformations at ground surface.
DEFORMATION MONITORING AT
GROUND SURFACE
5
Measurements of deformations at ground surface are

crucial in urban tunneling projects where damage to
surface structures and utilities should be prevented.
These measurements typically include settlements
(and heaves) of structures as well as tilting. Such measurements are performed with surveying instruments
(Precise Geodetic Level, total stations and GNSS), or
with geotechnical instruments like Electronic Liquid
Level Gauges, ElectrolyticTilt Sensors (electro-levels)
surface clinometers/tilt meters, precise taping, and
crack-meters.
Precise leveling and faade monitoring are the most
common methods for monitoring displacements at
ground surface. The accuracy of these measurements
DEFORMATION MONITORING IN
THE GROUND
These measurements record the deformation of target

positions in the ground, either around the tunnel or
deep below the ground surface. They are often used to
calculate strains or control the deformation of characteristic points in the ground (e.g. below the foundation
of buildings, below utility lines, etc.) (Sakurai 1981).
Such measurements are performed with geotechnical instruments including single- or multi-point
borehole rod extensometers, magnetic extensometers, sliding micrometers, inclinometers, probe deflect
meters (often called sliding curve meters) and deep
settlement plates. These instruments can be installed
429
either from the ground surface (before the tunnel face

reaches the area of the instrument) or from inside
the tunnel (radically from the tunnel wall or along
the tunnel axis ahead of the excavation face).
DEFORMATION MONITORING IN
THE TUNNEL
Tunnel wall convergence (closure) between references

points (hooks) bolted on the tunnel walls is usually
measured with standard metal tape extensometers.
For distances up to about 1015 meters, the accuracy of such measurements is typically 0.2 mm. The
method is easy to use and maintain but it only offers
the magnitude of the deformation along the line of
measurement.
Because of this disadvantage, in most present-day
tunneling applications, deformations of the tunnel
walls are obtained in three dimensions, by routine
geodetic surveying using total stations with integrated
distance measurement. In such applications, optical
reflector targets are installed at regular distances along
the tunnel axis (e.g. at sections every 1520 m) and,
on each section, at selected locations of the tunnel
wall (e.g. five reflectors per section: at the crest, at 45
degrees and at the spring-line). As tunnels are usually
long, the fixed (stable) reference positions are typically
located outside the tunnel, often at distances exceeding one kilometer and usually out of sight from inside
the tunnel.
Thus, measurements of the targets inside the tunnel
are obtained by placing the total station at pre-defined
rugged stations1 (bolted on the tunnel wall) and successively moving the instrument forward (towards the
tunnel excavation face) while measuring the coordinates of the visible targets from each station. The
theoretical accuracy of these measurements over a distance of about 100 m) is about 23 mm for lengths
and 2 for angles. For long tunnels, the accuracy
of the measurements is usually reduced by unclean
atmosphere and due to the multiple positions of the
instrument (especially if these positions are not stable
due to creep deformations of the tunnel).
A recent development of measuring the geometry of tunnel walls in cross section (and thus assess
the deformation in the interval between two measurement epochs) is the Tunnel Profile Scanners (profile
meters). In addition to measuring tunnel wall convergences in time, profile meters are also employed for
a variety of other purposes like comparing the actually excavated tunnel cross section with the design
requirement and for measuring the volume of shotcrete
placed on the excavated rock surface (by measuring
the profile before and after shotcreting).Tunnel profile
meters are fully digitized photogrammetric measuring
devices.
Figure 1. An example of settlement monitoring result (Villy

Kontogianni et al. 1999).
A typical such system consists of two CCD cameras

which are mounted on a portable frame. The cameras produce stereoscopic digital images of the tunnel
surface. The position of the camera frame is automatically determined by a total station with automatic
target recognition placed up to a maximum distance
of 100 meters.
For this purpose, three reflector targets are permanently mounted on the frame. Digital images are
automatically stored in a laptop computer and can be
processed to provide the 3-D coordinates of the surveyed tunnel wall surface with an accuracy of 5 mm
for each coordinate.
Although the accuracy of this method is low compared to routine geodetic surveying, the advantage of
recording a very large number of points on the tunnel
wall outweighs the low accuracy in many applications.
6.1 Traditional geodetic measurements
Three dimensional coordinates of object points (deformation points) which will be constructed on ceiling,
sole, and side walls of cross- section determined at different intervals on tunnels were measured by electronic
instruments. The coordinates of object points which
were constructed on ceiling, sole, and side walls were
measured at determined period intervals. By using
coordinates measured at different periods, forming
movements were determined. (Figure 1 is an example
of settlement monitoring result).
6.2 Automatic deformation monitoring
6.2.1 Automatic deformation monitoring
requirement
High-quality precision optical monitoring targets
mounted on support of excavation and tunnel linings, with fully automated motorized total stations
under computer control to monitor remotely the three
components of movement. Also required was a measurement precision of 1 mm for sight distances up to
100 m, with wireless data links to the control site. The
430
robotic total stations would have to be totally automatic and operate unattended 24/7 under all weather
conditions, and also be insensitive to refraction effects
caused by temperature or pressure variations. Each
target would have to be hit at least every 30 minutes.
Multiply TCA2003s, located where they could
monitor hundreds of target prisms to be affixed at predefined intervals to the structures to be monitored.
The measurement data would be transmitted at specified intervals via radio data links to a central location,
situated miles from the actual instrument monitoring stations. The total network would be tied together
and controlled by GeoMoS (Geodetic Monitoring
Software) system.
6.2.2 Installation challenges
There were two TCA2003 total station sites on each
side, for a total of four. The total stations were permanently mounted on fixed location with vented
heavy-duty glass enclosures to protect the system from
the elements. Each installation included an intercom
radio and modem with directional antenna to transmit
data from the site to the controller. Band pass filters
were added to overcome the high levels of RF activity
in the airport environment. The pedestals were isolated
to eliminate vibration and movement.
The GeoMoS software was installed on the computer network, and was configured so that data can be
accessed via a secure IP link by authorized personnel
and the resident engineer. Protecting the remote sites
from wind and weather presented special challenges.
The glass enclosures had to be robust and rugged, capable of withstanding heavy winds, rain, snow and ice,
and they had to be non-reflective so as not to distort the
EDM signal. Necessity being the mother of invention,
they experimented with a number of different types of
enclosures including clear round dog igloos until they
found the right solution.
6.2.3 Monitoring software
The GeoMoS software is a powerful tool for controlling the network of the remote sites, as well as
collecting data, providing alarms, post-processing,
reporting and visualizing data. The software represents
the data and results in graphical or numerical format.
You can select a time-line graph showing the trends of
movement over selected time periods (Burland 2001).
Multiple points can be viewed simultaneously in the
same graph. Alternatively, you can select a vector view
that shows displacement for a selected area, to easily
see where the greatest movement has occurred.
Senior-level project engineers and other authorized
personnel can access the GeoMoS data through a realtime web-based portal from any PC or laptop. They
can log onto the secure GeoMoS site and download
system status and reports. Measurement tolerances are
established and loaded into the GeoMoS system, If any
Figure 2. An example of tabular report of Automatic Deformation Monitoring for Hong Kong KSL Railway (Tang et al.
2007).
of these tolerances are exceeded, an automatic alarm

is activated (Figure 2 is an example of tabular report).
7
CONCLUSION
In order to determine deformations of the subway tunnel in soft ground, deformation monitoring must be
taken on the ground surface, in the ground and in the
tunnel. Practice shows that the traditional deformation
measurement can be achieved with 0.51 mm displacement accuracy, but deformation measurements
can also be recorded, processed and evaluated in
real-time by digital recording and telecommunication
systems. Real-time processing of ground deformations
offers the possibility of rapid response to upcoming
situations but requires advanced technology and an
appreciably higher cost. So real-time monitoring is
limited to cases where rapid response is absolutely
necessary, i.e., mainly in urban tunnels near sensitive
structures.
REFERENCES
Burland, J. B. & Standing, J. R. 2001. Building Response to
Tunneling Case Studies from Construction of the Jubilee
Line Extension. London: Thomas Telford publishers.
Clough, R.W. 1960. The Finite Element Method in Plane
Stress Analysis, Proceedings of American Society of Civil
Engineers, 2nd Conference on Electronic Computations.
23: 345378.
Dunnicliff, J. 1993. Geotechnical Instrumentation for
Monitoring Field Performance. U.S.A: John Wiley &
Sons Inc.
Hisatake, M. 1999. Direct Estimation of Initial Stresses of the
Ground Around a Tunnel. Proceedings of the Numerical
Methods in Geomechanics: 373377.
Kaiser, P. 1993. Deformation Monitoring for Stability
Assessment of Underground Openings.Compressive Rock
Engineering: 607630.
Kavvadas, M. 1999. Experiences from the construction of the
Athens Metro project, Proceedings of 12th European Conference of Soil Mechanics and Geotechnical Engineering:
16651676.
431
Kontogianni, V. & Stiros, S. 2002. Shallow Tunnel Convergence Predictions and Observations. Engineering
Geology 63(34):333345.
Kontogianni, V. & Stiros, S. 2003. Tunnel Monitorig During
the Excavation Phase:3-D Kinematic Analysis Based on
Geodetic Data. Proceedings of 11th FIG Symposium on
Deformation Measurements. Santorini, Greece.
Kontogianni, V., Tesseris, D. & Stiros S. 1999. Efficiency of
geodetic data to control tunnel deformation. Proceedings
of The 9th FIG International Symposium on Deformation
Measurements. Olsztyn, Poland:206214.
Liu, S. 2006. Deformation Measurements During the Construction of Large Dam Projects. Chinese Journal of
Underground Space and Engineering 06(Z2):13461348.
Liu, S. & Zhao, Z. 2007. Deformation Monitoring of 70 m
Span Box Girders of Hang-Zhou Bay Sea-Cross Bridge at
Construction Stage. World Bridge 07(2): 5860.
Mihalis, I. & Kavvadas, M. 1999. Ground Movements Caused

by TBM Tunnelling in the Athens Metro Project. Proc.
Int. Symp. on the Geotechnical Aspects of Underground
Construction in Soft Ground, Tokyo, Japan, June 1999:
269274.
Moaveni, S. 2003. Finite Element Analysis. New Jersey:
Pearson Education.
Sakurai, S. 1981. Interpretation of Displacement Measurements. Proceeding of the International Symposium on
Weak Rock, Tokyo: 751756.
Tang, E., Lui, V. & Wong, A. 2007. Application of Automatic Deformation Monitoring System for Hong Kong
KSL Railway. Monitoring Strategic Integration of Surveying Services. Proceedings of FIG Working Week 2007,
Hong Kong SAR, China.
432
The construction and field monitoring of a deep excavation in soft soils

T. Liu & G.B Liu
Tongji University, Shanghai , P.R. China
C.W.W. Ng
Department of Civil Engineering, the Hong Kong University of Science and Technology, HKSAR
ABSTRACT: This paper describes the construction of a 40 m multi-propped deep excavation in the downtown
area of Shanghai, China and the interpretation of the monitoring data. The entire excavation was 263 m in length,
23 m wide and 3841 m deep. The excavation was divided in three sections, i.e., the eastern, the middle, and the
western pits. Noticeable characteristic of this project are the excavation of a nearly 60 m wide section embedded
in the Huangpu River and the presence of many sensitive buildings nearby. In this paper, some construction
details of the excavation and the monitoring results are presented and discussed.
1
1.1
INTRODUCTION
Excavations for underground railway networks
in Shanghai
With the rapid development of Shanghai and the

citys underground railway networks, metro construction has become a multidisciplinary and multifaceted
project. During construction, regular service of neighboring underground railway lines must remain open.
Figure 1 shows the distribution of completed and ongoing excavations for underground railway networks
in Shanghai as of 2007. 92% of the metro stations are
deeper than 15 m and 31% are over 20 m in depth.
Zhao & Yang (2004) have suggested that any excavation, which is over 20 m in depth, will impose
severe geotechnical challenges in soft soils. This paper
describes the construction of a 40 m deep excavation in
the downtown area of Shanghai, China and the analysis
of its construction impacts to the environment.
1.2 Engineering background
The tunnel between the South Pudong Road Station
and the Nanpu Bridge Station is a cross-river section on
Line 4 of Shanghais underground railway system (see
Figure 2). On 1 July 2003, a failure occurred resulting
in flooding of the tunnel from a sub-artesian aquifer.
This led to the collapse of some sections of the tunnel,
excessive ground settlements and the closure nearby
buildings because of the potential danger to the public.
Figure 1. Statistics of depth of completed and on-going

excavations for underground railway networks in Shanghai.
An in-situ restoration program was implemented in

which an open cut was made in the damaged sections
of the tunnel to expose the collapsed sections, clear the
debris, and then to construct a new open structure.
As the extent of the collapse was large, the restoration work was constructed both on shore and in the
river. Using a cofferdam, a piled steel platform was
constructed above the river to excavate the collapsed
tunnel after it had been backfilled. Two undamaged
tunnel sections, which were flooded, were de-watered
by high-pressure pumps and then reinforced. The
entire restoration project included five major components: three deep excavations at the eastern, middle,
and western parts of the tunnel, clearance work for
433
Figure 4. Ground profile at the restoration project site.

Figure 2. Line 4 of Shanghais underground railway.
Figure 3. General layout of the restoration project.
the undamaged tunnel in Pudong (1003 m) and in Puxi

(760 m), and the connection work at the two ends of the
damaged and undamaged tunnel sections. The general
layout of the restoration project is shown in Figure 3.
A 65 m deep 1.2 m thick concrete diaphragm wall
was used to retain the three excavations (or so-called
pits). The total length of the excavations was 263 m
in length and 23 m wide (see Figure 3). The depth
of the excavations varied from 38 m to 41 m. Ten and
nine levels of concrete propping slabs were installed
to support the diaphragm wall at the eastern and the
western pits, respectively (see Figure 4). In the middle
pit, 35 m long, 1.2 m in diameter cast-in-place piles
were constructed to support the steel columns whose
cross-sectional area is 650 650 mm2 . In turn, these
columns were used to carry the horizontal propping
slabs.
2
Figure 5. Comparison of strata on site.
collapse. The physical and mechanical properties of

the soil layers are summarized in Table 1.
Based on the ground investigations, groundwater
was expected along and above the tunnel. The groundwater table is 0.51.0 m below the ground surface. The
ground comprises clay and silty clay layers at shallow
depths and more sandy materials at greater depths.
Layer 7 is the first aquifer in Shanghai, while layer 9
the second aquifer. They are connected hydraulically
on site but apparently they are not connected to the
Huangpu River. The groundwater and the Huangpu
River have no obvious hydraulic link on site.
ENGINEERING DIFFICULTIES AND

CONSTRUCTION TECHNOLOGY
GEOLOGICAL CONDITIONS
The restoration project was located adjacent to Outer

River Road, Dongjiadu Road and Zhongshan South
Road, where the terrain was quite flat. Figure 5 compares the current geological profile obtained from a
shift put down after the tunnel had collapsed and the
previously surveyed data obtained before the tunnel
The scale of the restoration work was large and the

depth of excavations was the deepest in Shanghai at
that time. Around the excavation site, there were sensitive buildings nearby such as the Linjiang Garden
Building and the Nanpu Bridge which required protections. Ground conditions were very complex owing
to the tunnel collapse and the materials left from the
434
Table 1.
Physical and mechanical properties of the soils at the restoration project site.
Layer
No.
Soil
Type
Thickness
(m)
Depth below
ground (m)
Water
content,
w (%)
Unit
weight,
(kN/m3 )
Void
ratio, e
Cohesion,
(kPa)
Friction
angle ( )
Average
SPT N
values
2
1&2
1
2
Fill
Silt clay
Gray clay
Green silty clay
Silty sand
Silty fine sand
Silty fine sand
7.70
1.8
7.10
3.5
12.30
22.86
4.29
15.09
22.19
25.69
37.99
60.85
33.5
30.9
43.3
24.4
35.1
28.2
25.2
18.7
18.7
18.2
20.2
20.3
19.8
0.98
0.90
1.24
0.70
1.04
0.78
0.71
14
7
14
43
0
0
0
27.5
32.0
12.5
15.5
33.0
37.0
35.5
10
40
50
50
rescue of the tunnel including slurry, polyurethane,

freezing agents, rails, and reinforced concrete pipes.
Since the depth of excavations reached the confined
water aquifer at layer 7 (see Figure 4), dewatering
was therefore required. As the pumping rate of a single well from dewatering tests was found to be about
90 m3 /h, the loss of groundwater was expected to high
during pumping. Special measures were taken to minimize ground settlements and hence to protect the
surrounding buildings. Several important construction processes of this restoration project are described
below.
3.1 Treatment of underground materials
The restoration work covered the main areas affected
by the tunnel collapse. Within the collapsed areas,
there were materials from building foundations and
abandoned underground pipelines, emergency rescue
materials, construction waste, sand backfill, previously abandoned buildings (e.g. the Wenmiao Pumping Station), and the septic tanks of the Linjiang
Garden building. In the deep strata, there were two
refrigeration units, tracks, sleepers, tunnel reinforced
steel supports, a large number of water pipes and the
collapsed reinforced concrete tunnel debris.
Materials at 1015 m depth were excavated and
removed using a heavy plant. The materials further
below were removed from excavation pits. Cutting
equipment was used to remove any obstructions, which
were taken away with the earthworks. When deep
underground obstructions affected the construction of
the diaphragm wall, a boring machine was used to clear
the obstructions (see Figures 6 and 7). The diaphragm
wall was then constructed after backfilling.
3.2
Construction of the diaphragm wall
To retain the deep excavation pits, which varied from

38 m to 41 m, a record depth of 65 m and 1.2 m thick
concrete diaphragm wall in Shanghai was required.
The construction of such a deep wall was expected to
encounter many difficulties. The available equipment
Figure 6. Cutting of the former tunnel by boring machine.
Figure 7. Barrier removal at the position of the diaphragm

wall.
and technology had to be improved for the construction. As shown in Figure 8, the two-drill-one grab
method was used. The method involves drilling two
oriented holes first with a rotary drilling machine and
then grabbing the soil between the holes effectively.
Using this method, the verticality of the holes was
better than 1/300 and the slurry trench excavation in
435
Figure 8. The two-drill-one grab method for the construction of the diaphragm wall.
Figure 10. The cofferdam and steel platform in the Huangpu

River.
Figure 9. Brushing work for the diaphragm wall.
the strata with obstructions could be carried out much

faster.
To achieve better efficiency in cleaning the soil at
the previously installed diaphragm wall panel, a counterweight was placed at the bottom of the diaphragm
wall trench (see Figure 9). This generated horizontal forces that made the brushing machine cling to
the joints by its directional bearings. In addition, the
method enhanced the waterproofing performance of
the wall.
3.3
Cofferdam and steel platform
The cofferdam and the steel platform in the Huangpu

River were important temporary structures, which contributed to the success of the restoration work in the
river (Figure 10). The foundation of the platform was
supported by 0.8 m diameter of steel pipes with a wall
thickness of 12 mm and 1.2 m in diameter cast-in-place
piles. The superstructure consisted of H700 steel pile
cap beams, H700 steel stringers, 18# I beam bridge
plan distribution beams, 10 mm bridge deck steel, and
45 steel pipe grates. The platform was +3.5 m above

ground and its width was 89 m.
3.4
Strengthening of the ground by jet grouting
In order to reduce the deflection of the diaphragm wall

during excavation and construction of the concrete
propping slabs, rotary jet grouting (see Figure 11) was
carried out below each level of slab, starting from the
fourth propping level and around the excavation.
Figure 11. Jet grouting construction.
The grout formed a frame-like earth beam to

enhance the shear strength of the soil in the passive
zone. Jet grouting was also carried out at the bottom of the excavation. The grouting was combined
with dewatering to protect the excavation from any
adverse effects such as inflow of water or sand and
base heave. Two jet-grouted piles were installed at each
joint between diaphragm wall panels to minimize the
ingress of ground water into the excavation (see Fig. 7).
3.5 Dewatering and monitoring
The excavation depth of the two open cut pits was
more than 38 m, which reached the aquifer of layer 7.
Dewatering became one of the most vital activities on
site. Since the dewatering was at great depth, ground
settlement had to be strictly controlled to protect the
neighboring buildings such as the Linjiang Garden
and the ramp of the Nanpu Bridge from any damage. Based on an optimized design using results from
in-situ dewatering tests, 56 wells of 61 m deep were
sunk and recharge wells were installed at appropriate
positions outside the pits. In addition to monitoring the ground settlements, water level observation
436
Figure 12. Dewatering monitoring points of the eastern pit.
wells, sub-surface settlement monitoring points, and

pore pressure observation holes were installed (see
Figure 12).
Figure 13. Plan showing the instrumentation installed on
site.
3.6 Three excavation pits

The excavation was 41.2 m deep at the two end sections and 38 m at the middle section (see Figure 4).
The pit was 22.3 m wide at the middle section and
23.7 m wide at the two end sections. Due to constraints
imposed by the presence of the deep obstructions, a
partition wall for dividing the entire excavation, which
was 174.1 m long in the eastern pit, and 62.5 m in
the western pit, could not be built. Because it was
a very long and narrow excavation, the stability of
the underground supporting system was very important. A three-dimensional structural supporting frame
with high rigidity, formed by upright column piles,
reinforced concrete supports and purlins, was used
to provide the required stiffness to control the lateral
deformations.
4
INTERPRETATION OF MONITORING DATA
Excavation for the eastern pit began on 1 March 2006

whereas that for the western and middle pits on 6 October 2006.The lateral deflections of the diaphragm wall,
settlement of the surrounding ground and buildings,
and bottom heave at each stage of the excavation were
measured at the monitoring points installed before
excavation (see Figure 13). Due to the page limit, only
selected data are presented.
Figure 14. Deflecion profile of the diaphragm wall (at
inclinometer I12) at each stage of excavation at the eastern pit.
4.1 Lateral deformations of diaphragm wall

Thirty-nine inclinometers were placed in the diaphragm
wall to measure its lateral deflections during excavation. A significant increase in the measured deflection
of the diaphragm wall was used as a warning signal.
Figure 14 shows the measured wall profiles by the
inclinometer (I12) at some key stages of excavation in
the middle zone of the eastern pit.
As shown in Figure 14, the measured deflection

profiles of the wall are typical for a multi-propped
excavation in Shanghai (Liu et al. 2005; Wang et al.
2005). As the excavation went deeper, the magnitude
of inward wall deflection increased as expected. The
largest deflection of the wall was 49.6 mm recorded
at a depth of 28 m. Figure 15 compares the measured
437
Settlement (mm)
-60
3th excavation
-65
5th excavation
Construction of
bottom slab
7th excavation
9th excavation
-70
4th excavation
6th excavation
8th excavation
Final excavation
-75
0
50
100
150
200
250
300
Elapsedtimefrom1-1-2006 (days)
Figure 15. Comparisons of the largest wall deflections

measured by all inclinometers.
Figure 17. The measured settlement at J5-2 of the Gutai

Restaurant versus time.
Gu
tai
Linjiang Garden
Re
sta
ur
an
Figure 16. The largest settlement values measured at all the

monitoring points installed at Gutai Restaurant.
Figure 18. The largest settlement values measured at Linjiang Garden.
maximum deflections at all inclinometers. The measured maximum deflections ranged from 20 mm to
50 mm, which are fairly small for a 40 m deep excavation as compared with other shallower excavations
in soft clays in Shanghai (Wang et al. 2005). This was
because of the large rigidity of the retaining system
provided and the early installation of props prior to
each stage of excavation.
-15
6th excavation
Settlement (mm)
-20
-25
9th excavation
4th excavation
7th excavation
5th excavation
-30
-35
8th excavation
-40
0
4.2
50
100
150
200
250
300
Elapsed time from 1-1-2006 (days)
Settlement monitoring
The adjacent buildings and structures around the excavation such as the Gutai Restaurant, Linjiang Garden
and the ramp of the Nanpu Bridge (see Figure 3)
required major protective measures and their settlement had to be monitored. Figure 16 shows the largest
settlement values measured at the Gutai restaurant. It
can be seen that largest recorded settlement of the
restaurant was 72.1 mm at the corner J5-2. This was
because dewatering was carried out near this corner
region. Figure 17 shows the measured settlement at
J5-2 versus time. As expected, the measured settlement increased as the excavation progressed. However,
upward movement (heave) was recorded after the 9th
stage of excavation and the rate of upward movement seemed to accelerate after the construction of
Figure 19. The measured settlement at J6 of the Linjiang

Garden versus time.
the bottom slab. This was probably of the increase in

pore water pressure and hence a reduction in effective stress due to the recharging of groundwater after
dewatering.
Figure 18 shows the measured maximum settlement values at the Linjiang Garden. As expected, the
measured maximum settlement of 38.9 mm was at the
corner closet to the excavation (i.e. at J6). Figure 19
shows the measured settlement at J5-2 versus time.
Similar to that shown in Figure 17, soil swelling was
recorded by J6 after the 8th stage of excavation but the
rate was smaller than that recorded by J5-2.
438
This was probably because the duration of monitoring

at this point was not long enough after the completion
of the western pit.
5
Settlement (mm)
Figure 20. The largest settlement values measured at the

ramp.
The diaphragm wall was being
installed at western pit and the
eastern pit was being excavated.
The western pit was

being excavated.
CONCLUSIONS
This paper provides a case history which illustrates

that a complex multi-propped deep excavation in saturated soft soils can be effectively engineered by proper
design and construction. The maximum lateral deflection (49.6 mm) of the 1.2 m thick diaphragm wall is
relatively small for a 40 m deep excavation in soft clays
in Shanghai.
ACKNOWLEDEMENTS
The authors would like to thank many colleagues
who have contributed to the construction and field
monitoring of the project and to acknowledge the
earmarked research grant 618006 provided by the
Research Grants Council of the Hong Kong Special
Administrative Region.
Elapsed time from 1-1-2006 (days)
Figure 21. The measured settlement at the middle of the

ramp (S4) against time.
REFERENCES
Figures 20 and 21 show the measured maximum

settlement at all monitoring points at the ramp of the
Nanpu Bridge and the settlement history of S4, respectively. The measured maximum settlement ranged
from 2.2 mm at S9 to 29.7 mm at S4, which was located
at the middle of the ramp. As expected, the measured
settlement along the ramp closer to the west excavation
pit was much larger than the values recorded away from
the pit. However, no soil swelling was recorded at S4.
Liu, G.B., Ng, C.W.W. & Wang, Z.W. 2005. Observed performance of a deep multi-strutted excavation in Shanghai
soft clays. Journal of Geotechnical and Geoenvironmental
Engineering, ASCE 131(8): 10041013.
Wang, Z.W., Ng, C.W.W. & Liu, G.B. 2005. Characteristics of
wall deflections and ground surface settlements in Shanghai. Canadian Geotechnical Journal 42(5): 12431254.
Zhao, X.H. & Yang, G.X. 2004. Practice and theory for
specially big and deep excavation engineering. China
Communication Press: 152156.
439
Excavation entirely on subway tunnels in the central area of the Peoples

Square
Y.B. Mei
Shanghai No. 7 Construction Co. Ltd., P.R. China
X.H. Jiang, Y.M. Zhu & H.C. Qiao

Shanghai No. 1 Construction Co. Ltd., P.R. China
ABSTRACT: The Open and Go-down square is located at the center of the city where the constructional
surroundings are very complex. Its a great challenge to the excavation because metro line 1 and 2 tunnels just
underlie the pit forming a shape of double crossing #. The detailed excavation process and many technical
measures are stated in the paper which may be referred by coming similar projects.
INTRODUCTION
With the rapid development of the state economy and

acceleration of urbanization, many domestic big cities
are facing problems such as land limitation, population expansion, traffic jam, environment pollution et
al. which restrict the continuable development of the
cities. To exploit the underground space can find new
space for the continuable development of the city, and
to develop the rail traffic is an effective way for solving
the problem of traffic jam.
With the large-scale exploitation of the underground space and formation of the city rail traffic
network, more and more deep excavations will be
very close to the existing rail traffic facilities, including 1) sharing the same retaining wall with existing
subway stations, 2) excavation at the side of running
subway tunnels, 3) excavation above running subway
tunnels, 4) excavation between the columns of the elevated light-rail, and all these cases are challenges to
the new excavation engineering.
GENERAL CONDITION OF THE CASE

PROJECT
crossing #, and the minimum distance from the bottom of the pit to the top of existing tunnels of metro
line 1 is only 3.3 meter. The features of this project are
mainly as follows:
1. The site is at the center area of the city, so the surrounding traffic is very busy, and the organization
of constructional vehicles is difficult.
2. Environment protection is very strict for the control
of the noise, vibration, dust, and wastewater caused
by construction activities.
3. Time limitation is critical because many construction activities can not be carried out before the
underground rail traffic stopped running at midnight.
4. For the safety of the running metro lines, the deformation of the underlying subway tunnels should
be controlled within certain scope, which makes it
very difficult for the construction.
5. The deformation of the nearby roads and underground pipelines should be strictly controlled.
6 Many concrete obstructions underlie and should be
demolished first.
3
The Open and Go-down square is located at the crossing point of West Nanjing Road and Middle Xizang
Road, neighboring New World Mansion and World
Trade Building, and its one part of the key project the
pivotal rail traffic project of the Peoples Square. The
area of the excavation is about 3300 square meter with
depth about 4 meter. The tunnels of running metro line
1 and line 2 underlie the pit forming a shape of double
GEOLOGICAL CONDITION
As well-known, Shanghai is one of the classical areas

that have deep soft soil which has poor mechanical
characteristics such as high compression, large deformation and obvious rheology. Therefore, the soft soil
is easily disturbed and excavations shall display sharp
Time and Space Effect.
According to the geotechnical report, within the
scope that excavation affects, the soil layers are mainly
441
The groundwater belongs to phreatic water type and

is supplied mainly by rainfall. The water table varies
from 0.5 m to 1.2 m below the ground surface.
The area of Open and Go-down square is about 3300

square meter with length of 66 m, 55 m in longitudinal
and latitudinal direction respectively. The slab of Open
and Go-down square ramps from ground surface to
3.8 m below in west-east direction.
For the excavation is not deep, a dam formed by soilcement mixing piles is adopted as temporary retaining
structure which has a perimeter of 137 m, width of
3.2 m. The length of soil-cement mixing piles varies
from 6 m to 11 m, and its strength should be no less
than 1.2 MPa. To reduce the impact on the subway
tunnels, high pressure jet grouting piles with same
parameters are used to substitute for soil-cement piles
above and between metro line 1 tunnels. The jet grouting piles have diameter of 800 mm, overlapped by
150 mm.
The life of subway facilities is 100 years for its
essentiality, and its safety during excavation is the key
point of this project. The deformation requirements of
station structures and tunnels are regulated as below
by metro-managing unit.
Figure 1. Location of Open and Go-down square.
Figure 2. Plan of retaining structures.

Table 1.
Soil layers.
Soil layer
Description
Thickness
m
1
2
1
3
4
2
Fill
Fill
Silty clay
Muddy and silty clay
Muddy clay
Clay
Silty clay
Silty clay
Fine sand
1.2
0.6
1.1
3.2
7.4
8.5
10.0
1.9
12.3
RETAINING SYSTEM AND REQUIREMENT

OF DEFORMATION CONTROL
composed by muddy silty clay and muddy clay with

thin layer of fine sand between them. Because of some
certain geological reasons, the soil layers and 1 are
absent. The soil layers from top to down are listed in
table 1.
1. Differential deformation of subway rails in

transversal direction should be less than 2 mm,
2. Differential deformation of subway rails in longitudinal direction should be less than 2 mm per
10 m,
3. The radius of deformation curvature of subway
structures should be greater than 15000 m,
4. The relative deformation curvature of subway
structures should be less than 1/2500,
5. The final absolute settlement and displacement of
subway structures should be less than 10 mm,
6. The rate of structures deformation caused by
excavation should be less than 0.5 mm per day,
7. The width of newly generated crack should be less
than 0.2 mm,
8. The final accumulated settlement of station and
tunnels should be less than 10 mm,
9. The final accumulated displacement of station
retaining walls should be less than 2.5 mm,
10. The accumulated settlement of structures in longitudinal direction should be less than 4 mm per
10 m,
11. The distance change between subway rails should
be limited within the scope of 2 mm to +6 mm.
For the particularity of this project, although the
excavation depth is not large, the pit engineering is
still designated to the highest grade, so the control of
442
Figure 4. Plan of directional jet grouting piles.
Figure 3. Section of retaining structure.
deformation indices are rather strict than before which

are listed below,
1 The accumulated horizontal displacement of temporary retaining structure should be less than
0.1%H (H is excavation depth),
2 The settlement of ground nearby the pit should be
less than 0.14%H.
5 TECHNICAL MEASURES OF EXCAVATION

5.1 Treatment of soft soil
Because of its poor mechanical characteristics,
Shanghai soft soil makes it rather more difficult in
underground engineering than in other areas. For the
soft soil is easily disturbed and easily deforms, usually
certain ways are adopted to improve its poor mechanical characteristics. To reduce the lateral deformation
of retaining structure during excavation, jet grouting
piles with diameter of 1200 mm at spacing of 800 mm
are used for soil treatment. Nearly the soft soil in the pit
are all treated from the ground surface to 3 m below the
bottom, and the total area reaches 2880 m2 . Over the
metro 1 tunnels, the jet grouting piles extend from the
ground to 500 mm to the top of tunnels. For the treated
soil is hard to remove, so the requirements below and
above the pit bottom are different. The strength of
treated soil below and above the bottom should reach
1.2 MPa and 0.6 MPa respectively.
Usually, when the jet grouting piles is being processed, there is fairly high pressure acting on the
very close soils, thus those soils would be compressed
and disturbed. To avoid the high pressure acting on
the subway tunnels, semicircular other than circular jet grouting piles are used adjoining the metro
line 1 tunnels (see fig. 4). The semicircular piles are
formed by jet grouting only at opposite side of tunnels
which is called directional jet grouting technology, and
this technology is developed during the past years in
Shanghai.
Figure 5. Special anti-uplift structure.
5.2 Special anti-uplift structure

The existing running subway tunnels are in balance by
the soil below and over them. If part of the soil over
them is removed, then the balance state will be broken,
and there seems to have a force dragging the tunnels
upward, so the tunnels may uplift. To prevent the tunnels from uplifting, new balance must be established.
In this project, a new kind of anti-uplift structure is
invented which is indicted in figure 5. At each side of
tunnel, reinforced concrete (abbreviated as R.C. hereafter) bored piles with diameter of 600 mm are set up
close to the tunnel by 500 mm. The end of drilling pile
is only 1 m from the underlying metro line 2 tunnels
at crossing point. When a small area of soil over the
tunnel is removed, the R.C. slab in same area will be
constructed quickly. At the same time, rebars of bored
piles are anchored into slab and welded together with
its rebars. Thus, the uplift capacity of bored piles can
443
Figure 6. Excavation step 1.
balance the uplifting force acting on tunnels which

equals the weight of removed soil.
5.3
Dividing excavation into pieces
It is not difficult to be understood that more soil are

removed from the top of the tunnels at one time, the
uplifting displacement will be larger. To control the
uplifting displacement within certain scope, the pit is
divided into three parts which are divided into pieces
either, so the procedure of excavation is fairly complex.
The whole excavation includes tree steps corresponding to the tree parts. The numbers in figure 6 indicate
the sequence of excavation in step 1. Excavation over
the metro 1 tunnels is relatively shallower which is
arranged to be carried out first. Usually, the width
of each excavation piece is 3 m which is determined
by experience. After the soil is excavated to the bottom, the bedding cushion layer and R.C. slab will be
constructed on time. The main rebars in the slab of
different pieces would be connected by mechanical
connector pre-embedded in the concrete. Figure 7 and
figure 8 demonstrate excavation in step 2 and 3, and
the numbers do not indicate the excavation sequence.
5.4
Loading on the finished slab
Another measure is taken to assure the safety of subway tunnels during excavation. That is, one piece of
Figure 9. Loading on slab.
slab will be loaded with heavy materials soon after

being finished and reaching certain strength. The loading must be equivalent to the weight of soil removed
from the tunnels top.
6
INFORMATION-BASED MONITORING
It is well-known that information-based monitoring

is necessary during the construction in underground
engineering. Due to vagueness and variability of
mechanical characteristics of soft soil, it is impossible to predict the deformation of retaining structure
accurately during excavation, so monitoring is the only
measure that can see the state of safety. Figure 10
shows the layout of monitoring points of metro line
1 tunnels.
Figure 11 demonstrates the deformation curve of
metro line 1 tunnel (upper one) during the first excavation step. The accumulated maximum uplifting displacement of underlying tunnel is 3.05 mm after the
complement of excavation in step 1. It is obvious that
444
maximum uplifting displacement of underlying tunnel

is 6.47 mm in step 2 which includes displacement generated in step 1. Generally speaking, the accumulated
deformation is fairly small during the construction of
underground engineering.
7
With the rapid development of the urban rail traffic,

more and more excavations are to be restricted by existing underground subway facilities such as excavation
under, above or between subway tunnels et al. One case
excavation completely on the running subway tunnels
is introduced in detail in this paper. The technical measures employed during the process of excavation and
the successful experience can be referred by similar
projects in future.
Figure 10. Monitoring points of subway tunnels.

4.0
CONCLUSION
displacement
3.0
2.0
1.0
ACKNOWLEDGMENTS
0.0
-1.0
This research was supported by Science and

Technology Commission of Shanghai Municipality
(No. 062012002 and 07QB14019 )
Oct.
Nov.
18th
27th
2nd
1
J2
Oct.
SC
J1
J1
SC
5
J1
SC
3
J1
SC
SC
J1
1
J0
SC
7
SC
J0
SC
J0
SC
J0
SC
SC
J0
monitoring point
REFERENCES
Garske E., Kauer H., & Von Soos P. 1989. Excavation lining and foundation for the new Kreissparkasse building
above existing subway tunnels in Munich, Bauingenieur
(German), 64(11): 505511.
Lo K.Y & Ramsay J.A. 1991. Effect of construction on existing subway tunnels, Tunnelling and Underground Space
Technology (6)1: 287297.
Eisenstein Z. Dan, Martin Geoffrey R., & Parker Harvey.
1997. Challenges of tunneling for the LosAngeles subway,
ASCE Construction Congress Proceedings, Managing
Engineered Construction in Expanding Global Markets:
281289.
Ookado, & Nobuyuki. 1998. Large-scale excavation in
Tokyo for the construction of the Namboku subway line,
Geotechnical Special Publication 86:124143.
Zhu Z. F., Tao X. M., & Xie X. S. 2006. The influence and
control of deep excavation on deformation of operating
subway tunnels, Chinese Journal of Underground Space
and Engineering (1):128131.
Zhang J. L., Liu G. B., & Liu H. 2006. Study on construction
technology of excavation on subway tunnels, Construction
Technology (4): 8590.
Figure 11. Uplifting displacement of subway tunnels during

excavation step 1.
7.0
displacement
5.0
3.0
1.0
-1.0
-3.0
Jul.
Aug.
18th
26th
3rd
1
CJ
2
Jul.
CJ
1
X
CJ
1
X
CJ
1
3
CJ
1
X
11
CJ
X
CJ
0
X
CJ
0
X
CJ
0
X
CJ
0
X
CJ
0
monitoring point
Figure 12. Uplifting displacement of subway tunnels during

excavation step 2.
the deformation is fairly small and the safety of running tunnel is guaranteed. Figure 12 demonstrates the
deformation curve of metro line 1 tunnel (lower one)
during the second excavation step. The accumulated
445
The benefits of hybrid ground treatment in significantly reducing wall

movement: A Singapore case history
N.H. Osborne, C.C. Ng & C.K. Cheah
Land Transport Authority, Singapore
ABSTRACT: The first major use of hybrid ground treatment, Jet Mechanical Mixing (JMM), in Singapore,
was in the construction of the new Nicoll Highway Station (NCH) for the Circle Line Project (CCLP). This
system played a major role in the success of the project. The retaining wall design was a 1.5 m thick diaphragm
wall to provide support of the ground during the station construction and also as the permanent walls of the
station. The proposed construction scheme is that of top-down with the use of a 7 m thick JMM below the base
slab level, which comprises a large diameter deep soil mixing method comprising a central core combined with a
jet-grouted outer circumference to form a large diameter of improved soil mass. This paper aims to demonstrate
the effectiveness of the excavation support system and benefits of JMM in reducing wall movement of deep
excavation in soft ground with 35 m thick of Marine Clay.
1
1.1
INTRODUCTION
1.2 Ground condition
Background
The construction works for the original Nicoll Highway Station (NCH) on the Circle Line Project (CCLP)
was halted when a collapse of the cut and cover tunnels
leading to the station occurred in April 2004. Following the collapse, several options were studied for the
recommencement of the works. The option to realign
part of the project to avoid the collapsed site was
eventually adopted. As a consequence of this realignment, NCH was relocated approximately 100 m to the
south, as shown in Figure 1, with the station design
and excavation restarting afresh.
Figure 1. Tunnel Alignment Drawing.
At NCH, the ground consists of man-made fill, fluvial

sands, fluvial clay and the Marine Clay of the Kallang
formation, underlain by the Old Alluvium, as shown
in Figure 2. The thickness of the fill is typically 3 to
6 meters. Underlying the fill is a layer of fluvial sand.
Beneath this, it is the very soft to soft Marine Clay. The
thickness of the sand layer is 3 to 7 m. The depth of the
Marine Clay varies from 30 m to 40 m below ground
level.
Locally, the Marine Clay is separated by a layer
of laterally discontinuous fluvial deposits. The fluvial
sands found at NCH are typically described as loose
to medium dense gray sands or silty sands. The properties of the fluvial sands are described by Chu, et al.
(2000). The properties of the Singapore Marine Clay
and problems associated with it from a tunnelling and
deep excavations perspective have been well established in Singapore; see Tan (1972), Shirlaw & Copsey
(1987), Chang (1991) and Tanaka et al (2001), and
generally relate to its softness. The Marine Clay is
normally consolidated or slightly over-consolidated,
with a undrained shear strength (Cu) starting at about
20 kPa and increasing slowly with depth. The compression index is typically in the range of 0.6 to 1.0.
The permeability is low and is in the order of 109 to
1010 m/s. The Old Alluvium is typically described as
sandy silt or clayey silt. At depth the material is generally found to have some cementation. However much
of the cementation has been lost due to weathering at
shallow depth. The permeability of the Old Alluvium
447
Figure 2. Ground conditions at the new NCH Station.
Figure 4(a). JMM Machine.
Figure 4(b). Schematic diagram of the drilling rod showing

the mixing arm of the JMM machine.
Figure 3. Typical cross section of NCH Station.
depends on weathering and grain size distribution. It

typically ranges between 106 to 109 m/s.
1.3 Section details
The new NCH was designed as a top down structure
with a 1.5 m thick permanent diaphragm wall anchored
into the Old Alluvium strata. Toe level is at 60 m below
ground level. This also reduces the need for temporary
struts as the permanent concrete roof and concourse
slabs are constructed during excavation. The excavation depth is 20 m below ground level and the width
of excavation is 24 m. The cross section of the station
box is shown in Figure 3.
1.4 Hybrid type of ground treatment
Ground treatment underneath the base of the station is
often used to limit the wall deflection, act as a working platform and prevent uplifting of these soft clayey
soils. The application of ground treatment such as jet

grout piles (JGP) for deep excavations in Marine Clay
in Singapore has been presented by Page et al. (2006).
For this project, the ground treatment option was Jet
Mechanical Mixing (JMM), a hybrid of jet grouting
and deep soil mixing. A proprietary name, RASJET
is given to it by the specialist contractor from Japan,
Raito Kogyo. This was the first time such a system had
been used in large scale in Singapore.
JMM is a combination of soil mixing and jet grouting that produces overlapping columns with an internal
column of mixed soil by the auger and an external
column created by a slurry jet into the in-situ soil.
The process of forming the columns is similar to the
method of forming JGP columns with the addition of
dual and counter rotation mixing blades on the drill
rod to ensure intensive soil mixing. Figures 4(a) & (b)
show the JMM machines, the drilling rod and the mixing arm of the machine. The rod/auger had a large
diameter of 457 mm as compared to the traditional
JGP rod of 200 mm. The high stiffness of the drill
448
ensure that a high level of quality control. It should

be noted that a further benefit is gained during the
withdrawal the auger after completing the JMM slab.
As the auger is withdrawn, lower quantities of cement
are added and the ground is mixed. This creates a 1.6 m
diameter treated column all the way to the surface, as
shown in Figure 6. Consequently the strength of the
soil above the treated ground is significantly enhanced.
2
Figure 5. Typical layout of the JMM columns.
2.1
INITIAL DESIGN OF NCH STATION

TEMPORARY WORKS
Initial design assumption
The initial design of NCH was carried out using moderately conservative parameters for the original ground
and JMM. A conservative approach was adopted for
the initial design and therefore no additional strength
attributed to the soil cement mix above the JMM
layer was considered. The Undrained Shear Strength
(Cu) and Youngs Modulus (E) of the JMM layer was
assumed to be 300 kPa and 90 MPa respectively, based
upon traditional design parameters used for jet grouting in Singapore. Table 1 shows the parameters of the
ground and JMM used in the initial design.
2.2
Figure 6. Columns created during withdrawal.
rod contributes to a more accurate drilling verticality. Combined with the rod were the mixing blades
which created an inner mechanical soil mixing column
of 1.6 m diameter. A jet grout nozzle on the mixing
blade introduces cement slurry mix with pressurized
air into the soil and adds a further 0.6 m of jet grouting around the soil mixing column, creating a 2.8 m
column within the ground. These columns are then
designed with appropriate overlap to provide a full coverage of the treated areas. Figure 5 shows the layout
of the mechanical mixing part and jet grouting part of
the JMM column.
There are numerous advantages to this system (Page
et al. 2006 & Ueda et al. 2007), principally the benefits of mixing and grouting are experienced. From
the mechanical soil mixing, a known treated area is
assured and from the jet grouting, a sizeable overlap and penetration into any shadow areas close to the
retaining system is achieved.
To install a JMM column, the auger is first drilled to
the base level of the JMM column with water injection,
and withdrawn to the top level of the JMM column
with mechanical mixing without any injection. It then
descends with slurry injection and mechanical mixing
to form the internal soil mixing column up to base
level. After which, it ascends with jetting to form the
external jet grouting perimeter. The whole process is
automated and monitored real time by data loggers to
Geotechnical analysis
The geotechnical analysis of the excavation sequence

was done using a two-dimensional analysis of deformation and stability with PLAXIS (Version 8), a
geotechnical finite element program. The non-linear
and stress-dependent stress-strain properties of the
soils are modeled as elastic perfectly plastic using the
Mohr Coulomb model. The undrained behaviour of
clays and cohesive materials is simulated in PLAXIS
using Mohr-Coulomb soil model in terms of effective
stresses with undrained strength parameters (commonly known as Method B in Singapore). This was
done by specifying the effective Youngs Modulus (E )
and Poisson ratio ( ) with Cu under undrained setting. In this method of analysis, the Cu of the clays
and cohesive materials were capped at the input values given by the user. The input soil parameters were
based on the moderately conservative values shown in
Table 1.
Figure 7 shows the cross section of temporary works
for the excavation. The 1 m and 1.8 m diameter bored
piles were modeled as a fixed-end anchor with horizontal spacing of 12.325 m centre to centre to avoid
possible ill effect of soil-beam interaction in a 2-D
plane strain analysis. The equivalent length of the
anchor is taken as half the pile penetration length,
which is 16 m. Temporary steel plunge-in column is
modeled using the node-to-node anchor to simulate
the behavior of an axially loaded member to avoid possible ill effect of soil-beam interaction. Though large
449
Table 1.
Soil parameters and strength of treated soil used in the initial design.
Stratum
Standard
penetration Bulk
test
density,
SPT N
(kN/m3 )
Undrained
shear
strength,
Cu (kN/m2 )
Drained
shear
Friction
strain,
angle,

2

C (kN/m ) ( )
Fill
M
64
01
19.0
16.0
0
0
30
22
F2
11 5
19.0
22
F1
E
11 7
04
20.0
15.0
25
Design: 15 (0 to 10 m)
Worst Credible: 13 (0 to 10)
Design: 15 + 1.2 (z 10)
Worst Credible: 13 + 1.1
(z 10)
20 (0 to 10 m)
20 + 1.14 (z 10)
0
15 (0 to 10 m)
15 + 1.2 (z 10)
OA(W) N < 30
25
20.0
OA(SW-2)
35 6
30 < N < 50
OA(SW-1)
70 13
30 < N < 100
OA(CZ) N > 100 >100
JMM
0
0
20.0
Sandy: 0
Clayey: 10
Design: 5 N (max. 500 kPa) Sandy: 5
Worst Credible: 3 N
Clayey: 20
10
20.0
16
300
20.0
25
Elastic
modulus,
E (kN/m2 )
10000
Design: 300 Cu
Worst Credible:
200 Cu
Design: 300 Cu
Worst Credible: 200 Cu
30
1500 N
18
Design: 300 Cu
Worst Credible: 200 Cu
Sandy: 32 Design: 2000 N
Clayey: 25 Worst Credible: 1000 N
Sandy: 32
Clayey: 25
30
32
90000
Figure 7. Cross section of temporary works design.
voids are present at concourse level at reduced level of

90.92 m, the diaphragm wall panels are supported by
the concourse slab abutting the diaphragm wall acting
in-plane as a waler beam. This waler beam support is
modeled as a fixed-end anchor in lateral direction.
As shown in Figure 7, there were 2 layers of temporary steel struts above the permanent roof slab during
the excavation. The first and second layers of struts

(S1 & S2) were H-beam of size 414 405 232 at
spacing of 5.7 m. The first and second layers of struts
were installed at reduced levels of 102 m (1 m below
the ground level) and 99.87 m respectively. The struts
were modeled in the program as fixed end anchors
with effective length of 12 m, which is half of the strut
length between the two diaphragm walls support. The
pre-stressing on the struts was typically 50% of the
design strut load.
On comparing the predicted finite element model
and the actual maximum wall deflections measured
by in place inclinometers, a very significant difference of 60 mm of movement is observed. A maximum
measured deflection of 25 mm compared to a predicted 85 mm wall deflection is observed, as shown
in Figure 8. This is a very notable discrepancy. This
difference rapidly became apparent as the excavation progressed. Although it was evident that much
of this disparity could be attributed to the JMM, as
suggested by the strength and stiffness results from
the site testing of the JMM, back analysis was undertaken to quantify these differences. Although a change
in deflection from 0 mm to 20 mm was observed
over 5 m length of the diaphragm wall, no visible
crack was observed. The deflection of diaphragm
wall could be less. This could be due to an inherent problem of the inclinometer installed in steel
pipes in the diaphragm wall with cement bentonite
backfill.
450
Figure 9. Comparison of JMM strength used for initial

design and actual values from test data.
Figure 8. Predicted wall deflection of the original design

and the measured all deflection.
Figure 10. Comparison of JMM stiffness used for initial
design and actual values from test data.
3 ACTUAL STRENGTH PARAMETERS FOR

JMM AND SOIL CEMENT MIX ABOVE
3.1
JMM strength parameters
Post installation and prior to the commencement of

excavation, an extensive quality check on the strength
parameters of the JMM layer as well as the soil cement
mix above was carried out. Figure 9 shows the summary of the analysis of the test results from the 7 m
thick layer directly beneath the base slab. It shows
the comparison between the average strength of the
JMM layer and the parameters assumed in the original design. Also included is a strength factored by
a mass correction factor to account for any variation
caused during the construction process, for use later
in further analysis (refer to Section 4.1 for details
of correction factor). The average strength, Cu, of
the JMM layer is about 1845 kPa, with very consistent strengths achieved in the samples tested, ranging
from a lowest of 1150 kPa to the highest of 2370 kPa.
The average strength is more than five times higher
than that originally assumed value of 300 kPa in the
design.
3.2
Figure 11. Average parameters of JMM and Soil Mixing

above.
Soil cement mix strength parameters
Similar findings are experienced on analysing the

results from the stiffness testing, Figure 10 shows the
comparison of the stiffness of JMM from test results,
the original design value and the factored value taking
into account the mass correction factor. The average
stiffness of the JMM from test results is 572 MPa,
with a lower bound of 400 MPa and an upper bound of

700 MPa. Again the tested average stiffness is significantly higher than the 90 MPa assumed in the original
design. The various strength of the treated ground
with JMM and soil cement mix above the JMM is
summarized in Figure 11.
451
BACK ANALYSIS
4.2 Interpretation of the back analysis results
4.1 Accounting for better strength parameters

As it was evident in the early stages of the excavation that the retaining wall performance, in terms of
movement, was appreciably better than that design
prediction. The review of the JMM test results suggested that they were the most plausible reason for the
reduced movement. Back analysis of the performance
was undertaken for two main objectives: (i) to model
the performance of the cofferdam with the JMM modeled more accurately; (ii) to justify the omission of the
third temporary strut, which was originally proposed
between the concourse and base slab. In principal, the
analysis was first done through modifying the initial
finite element model using most probable parameters
to match the actual wall deflection and strut load of the
second stage of excavation. It was refined thereafter as
the excavation proceeds. At the time of the back analysis, the second stage of excavation was the current
excavation progress.
The most probable parameters include the most
probable soil parameters, JMM parameters and the
improved parameters of the soil mixing columns above
the JMM. The most likely parameters for 7 m thick
JMM slab and the soil mixing column above the JMM
slab were based on the actual core tests results multiplied by a factor (0.725), which coincides with most
probable or mean mass correction factor for the JMM.
The correction factor is applied to account for the
potential non-uniformity of the ground treatment. This
correction factor is derived from the estimated coefficient based on the square of the ratio of seismic
wave velocity, obtained from cross hole seismic testing (Vsf), to the seismic wave velocity obtained from
laboratory Ultrasonic test (Vsc), i.e.
The results from the various back analysis are summarized in Figure 12. The wall deflection from the
back analysis (Analysis 1) using the most probable
parameters at second stage excavation, with the mass
correction, is able to reasonably match the reading of
the in wall inclinometers at the same stage. This is
supported by the fact that the reading of strut force
at first level strut is quite similar to the strut force
obtained from back analysis. However, the wall deflection obtained from the Analysis 2 using unfactored
parameters gives a much closer match to the actual wall
movements. This suggests that the quality, strength and
stiffness of the JMM are consistent throughout the
entire JMM slab and no correction factor is needed
to be applied.
CONCLUSIONS
This case history demonstrates that JMM, if properly

installed, has major benefits in controlling the stability and movements induced by deep excavations in
soft ground. The reasons can be attributed to the fact
that the inner soil column is comprehensively mixed,
combined with the attributes of the outer jet grouted
column with sufficient overlapping.The whole process
undergoes tight quality control and rigorous testing to
ensure a continuous and comprehensive slab. In addition to the JMM slab, there is the major benefit of
Correction Factor = (Vsf/Vsc)2

The results of the wave velocities (Vsf and Vsc) and
the rational of deriving the correction factor from the
wave velocities is outlined in a paper by Ueda et al.
2007. Based on these parameters, back analysis was
carried out to match the actual wall deflection and strut
forces, as close as possible, up to second stage excavation. Table 2 shows the most probable parameters for
the JMM and soil above the JMM.
Table 2.
Figure 12. Wall deflection for the back analyses
Most probable JMM parameters for back analysis.
Most probable JMM parameters

(Correction Factor = 0.725)
Most probable parameters for soil cement

mix above the JMM (Correction Factor = 0.725)
94.6 m < RL < 102.5 m(GL)
81.3 m < RL < 94.6 m
Cu (kPa)
E50 (MPa)
0.725 178 = 129

0.725 24 = 18
0.725 1845 = 1337.6

0.725 572.1 = 414.8
452
0.725 339 = 245.8

0.725 44 = 31.9
the discrete soil mixing columns formed during the

withdrawal of the auger.
Based on the limited usage to date it is difficult
to suggest what parameters should be used for future
design. However it is clear that the key is in the
quality control of the process in ensuring a total and
uniform treatment. With todays engineering sophistication, this can be achieved. The strength, stiffness
and quality of the JMM is significantly higher than jet
grouting, but the choice of actual design parameters to
be used is complex. It is recommended that they should
be determined on a case by case basis with local trials
specific to the ground conditions, but considering the
strengths and stiffness already achieved. The benefits
of JMM should not be ignored and this technique will
be a future benefit to the industry in controlling ground
movements.
REFERENCES
Page, R.J., Ong, J.C.W, Osborne, N. & Shirlaw, J.N. 2006.

Jet Grouting for Excavations in Soft Clay Design and
Construction Issues. Proc. of International Conference on
Deep Excavations, 2006.
Shirlaw, J.N. & Copsey, J.P. 1987. Settlement over Tunnels
in Singapore Marine Clay. Proc. of the 5th International
Geotechnical Seminar Case Histories in Soft Clay,
NTI, Singapore, 1987.
Tan, S.B. 1972. Foundation Problems in Singapore Marine
Clay, Asian Building and Construction 1972.
Tan, S.B., Tan, S.L. & Chin, Y.K. 1985. A Braced Sheetpile Excavation in Soft Singapore Marine Clay. Proc.
of 11th Conference on Soil Mechanics and Foundation
Engineering, San Francisco, 1985.
Tanaka, H., Locat, J., Shibuya, S., Tan, T.S. and
Shiwakoti, D.R. 2001. Characterization of Singapore,
Bangkok and Ariake Clays, Can. Geot. J. 38:378400.
Ueda,Y., Furusone, T., Sato, Y. & Imamura, S. 2007.
Design, Construction and Quality Control of the Ground
Improvement Method Adopted for Singapore Marine
Clay Mechanical Soil Mixing with Cement Slurry Jet
Grouting. Proc. of 4th Civil Engineering Conference in
the Asian Region 2007.
Chang, M.F. 1991. The Stress History of Singapore marine

clay. J. Geotechnical Engineering, Vol. 22, 1991.
Chu, J., Wen, D., Kay, R.E. & Tay, T.H. 2000. Engineering
Properties of fluvial sands at Race Course Road. Proc. of
the International Conference on Tunnels & Underground
Structures, Singapore, 2000.
453
3D deformation monitoring of subway tunnel

D.W. Qiu, K.Q. Zhou & Y.H. Ding
Beijing University of Civil Engineering and Architecture, Beijing, P.R. China
Q.H. Liang & S.L. Yang

Beijing Jiaotong University, Beijing, P.R. China
ABSTRACT: As a kind of modern vehicle, subway has shown us the advantages of safety, speediness, low
power consumption, low pollution etc. It has been the main part of the urban high-capacity public traffic system.
More and more subway lines appear in order to meet the need of the society. The civil engineering construction
especially in Soft Ground must lead to the deformation of the adjacent subway tunnel, and it causes the severe
influences to the stabilization and safety of the tunnel. This paper puts forward a kind of three-dimension
deformation monitoring method. It can precisely monitor the deformation of the tunnel liner real-time which
do not interrupt the subway transport, and expediently provide mechanics analysis on the deformation of tunnel
construction. The paper discusses how to build the 3D mathematical model for the subway tunnel by the ground
lidar surveying technology. Use the Georobot to survey the deformation of the tunnel cross section and the rail
automatically. According to data processing and analysis using a model interpolation method, the 3D digital
model of the deformation and displacement for the whole metro is fitted finally. Thus we can obtain the 3D
image of the subway deformation in a precision, real-time, stereo and visual way.
INTRODUCTION
In order to alleviate the pressure of the ground transportation, urban subway has been programmed or built
in big cities. As a kind of modern transportation facilities, subway has shown us the advantages of safety,
speediness, low power consumption, low pollution etc.
It has been the main part of the Beijings high-capacity
public traffic system. Along with the development of
city in Beijing, more and more civil engineering construction must lead to the deformation of the adjacent
subway tunnel. It causes the severe influences to the
stabilization and safety of the tunnel, thus endanger the
whole urban transportation system. During the construction of periphery foundation ditch engineering
and tunnel engineering, how to guarantee the safety
of the subway tunnel has been the severe difficulty
needed to be solved.
The traditional monitoring method is to set monitoring cross sections on the deformation district of tunnel.
By surveying these monitoring points with convergence rule, total station and leveling, we can monitor
the deformation of the tunnel structure. Some drawbacks exist here. First, the number of the monitoring
points is limited, which can not reflect the deformation tendency. Load analysis of the deformation of
tunnel structure is constricted and the corresponding
reinforce measure is difficult to carry out. Second,

the traditional one has not used the remote monitor
method, so it disturbs the transportation of the subway to some extent. Third, the dim light, narrow space
and the complicate environment do disturb the safety
monitoring.
A method of three-dimension safety deformation
monitoring for urban subway is put forward here. It
can obtain the 3D digital data of the subway tunnel
deformation, which do not interrupt the subway transportation. It can not only precisely monitor the deformation of the tunnel structure and get the tendency
of subway deformation, but also provide mechanics
analysis on tunnel structure and the rail. This method
has been applied to the safety monitoring of the Beijing
subway line 1 (Babaoshan Station Bajiao amusement
park Station).
2 THREE-DIMENSION SAFETY MONITORING

METHOD
2.1
Model building
Lidar technology is also called three-dimension laser

scanning technology, which is a new kind of non-touch
surveying method. It can obtain the array geometric
455
Figure 1. Reflection target setting.
image of survey object from laser point clouds, which

were emitted by scan prism and the quick laser ambulator. So the three-dimension space model can be
made. This technology can obtain the three-dimension
coordinates of one point without reflecting prism,
and the speed can reach 100,000 points/second. This
technology is well suited for many applications: industrial, architectural, civil surveying, urban topography,
reverse engineering, archaeology.
The detailed monitoring project is introduced here:
2.2.1 Measurement principle of Georobot

Georobot is often called automatic electronic total
station (ETS). It is a kind of intelligent electronic total
station, which is able to search target automatically,
recognize, trace, collimate precisely and obtain the 3D
coordinates.
Target points observation of tunnel deformation
adopts free stationing principle of Georobot. Reflection sheets are set on the target points. To achieve
higher resolution and improve reliability of observation data, Georobot can finish redundant observation
automatically under the control of on-board software.
Then adjustment of observation data of different periods, 3D coordinates values in different periods will
be done by post-software of computer, finally we can
get 3D coordinate displacement of target points: (X,
Y, Z). For observation networks of free stationing,
we choose indirect adjustment model to process data.
Let t be necessary observation number and n be total
observation number (n > t). Then adjustment model is:
1. Within the deformation district, set the annular

closed survey control network along the middle line
of the up-down rail.
2. Put a set of cross section every 3 m following the
middle line of tunnel. Set the reflection target on the
arch, vertical wall, subgrade of the rail (As Fig. 1
shows), and collect the point clouds information by
three-dimension lidar scanner.
3. Based on the NURBS Curved face function (refer
with: Eq. 1), the three-dimension model of tunnel
is established by the data process such as data joint
and registering.
The corresponding error equation by the matrix is:
Let the order of polynomial is p q, NURBS

Curved face function can be written as:
The above-mentioned method is that gets displacement by comparing coordinates of observation points.
In the course of subway tunnel deformation analysis,
we also consider lateral vector after observation value
adjustment of different periods:
where di, j (i = 0,1, , m j = 0,1, , n) represent
the control vertex, and wi,j denotes the weight factor
of vertex, and Ni, p(u) , Nj, q(v) is gage B spline primary
function.
2.2
Real-time safety monitoring system
The 24 hour safety monitoring is necessary to guarantee the safety of both the tunnel structure and
transportation of the subway. Considering the high
density of the subways transportation, we adopt the
remote automatic monitor system to real-time monitor the tunnel structure, vertical wall, and subgrade of
the rail by the Georobot.
where, xi , yi , zi are adjustment value. k is observation

times of tunnel deformation, k1,k is the convergence of measurement lines, they do not include error
from possible displacement of base points, so they can
accurately reflect tunnel deformation. By regressive
analysis of these adjustment deformation values of different periods, we can conclude forecasting results of
tunnel structure deformation.
2.2.2 The setting of Georobot
The specially made instrument pier is put outside the
right of the first rail. Georobot is forced to be fixed
456
on the instrument pier through the pedestal and is

protected with the glass cover. 616 reflecting sheets
are installed for each monitoring station, which is distributed in the arch, vertical wall, orbit drainage ditch,
rail fastener and so on. With the monitor program,
Georobot collects the coordinates automatically and
transports the data to the control server via data wire.
3
ENGINEERING APPLICATIONS
The thermal pipeline engineering of Babaoshan south

road crosses over the structure of Beijing subway
line 1 (Babaoshan Station Bajiao amusement park
Station). Engineering construction causes the subway
tunnel deformation. The kilometer post of the tunnel
deformation district is from K3+770 to K3+810.
The safety monitoring result is shown as follows:
the accumulative deformation value of the max deformation point on the tunnel structure is +1.90 mm;
the max accumulative value of the rail deformation
is +1.86 mm; the max differential settlement of the
rail subgrade is 0.29 mm. So such conclusion can
be made that the accumulative displacements of both
tunnel structure and rail caused by the engineering
construction are less than 2 mm, which is within the
allowed deformation range and put no influences on
the subway transportation.
4
the safety monitoring of high-rise building, side slope,

deep foundation ditch engineering and so on. Of course
this method is not so mature. Its theory needs to be
further researched through engineering experience.
ACKNOWLEDGEMENTS
The authors wish to acknowledge the support of
the Key Technologies R&D Programme of China
(No. 2006BAJ15B01) and the Beijing University
of Civil Engineering and Architecture Science
Research Foundation of China (No. 100701205,
No. 100601905).
REFERENCES
Bassett, R.H., Kimmance, J.P. & Rasmussen, C. 1999. Automated electrolevel deformation monitoring system for
tunnels. Proceedings of the Institution of Civil Engineers,
Geotechnical Engineering: 117125.
Wu, Z.A. 1989. Processing Deformation Data of Engineering
Construction. Beijing: Surveying Publishing House.
Xu, W.P. 2000. Exploring on Feasibility of New Way to Monitor Tunnel Displacement. Railway Engineering Transaction 66(2).
Yu, L.F. & Duan, D.Q. 1996. Real-time Theodolite Industrial
Surveying System. Beijing: Surveying Publishing House.
CONCLUSIONS
This method of three-dimension safety monitoring has the advantages of high automation and
three-dimensional measurement. It can be applied to
457
Challenging urban tunnelling projects in soft soil conditions

H. Quick, J. Michael & S. Meissner
Prof. Dipl.-Ing. H. Quick, Ingenieure und Geologen GmbH, Darmstadt, Germany
U. Arslan
Univ.-Prof. Dr.-Ing. U. Arslan, Technische Universitt Darmstadt, Germany
ABSTRACT: Different challenging tunnel projects in the downtown area of the city of Mainz in Germany
are presented. These tunnels run parallel in a distance of 4 m to max. 50 m. The tunnels were built in soft soil
conditions consisting of filling, clay and marl layers of the Tertiary. The paper presents the different construction
techniques, the calculation methods for the two tunnels as well as the results of measurements for the New
Tunnel Mainz. The experience for the construction of this tunnel and the results of the measurements were the
basis for the chosen construction and calculation method for the rehabilitation of the Old Tunnel Mainz, which
is currently under construction.
INTRODUCTION
The New Tunnel Mainz had been constructed in the

years 1998 to 2001 directly adjacent to the existing Old
Tunnel Mainz. This old tunnel built in 1884 is currently
being rehabilitated, converted and enlarged during the
next years. Due to the small overburden of both tunnels
and sensitive structures on the ground surface challenging and unique tunnelling techniques were chosen
to guarantee the stability and serviceability of the tunnels and of sensitive structures. In addition calculation
methods and results from geotechnical measurements
are presented in the following. The situation of the
tunnels is shown in figure 1.
GROUND CONDITIONS
The geological condition is mainly characterized

by the tertiary strata (Miocene) of the Mainzer Basin.
The Tertiary strata sequence consists of an alternating sequence of marly clays, chalk marl, sandy silts
(hydrobia silts, hydrobia oyster shells) and sands in an
alternating sequence with chalkstone banks (fig. 2).
The chalkstone banks are partly compact/massive to
weathered. The consistency of the in-situ ground is
stiff to semi-solid, turning into soft/paste-like if water
intrudes. The groundwater can be found up to the level
of the floor/upper edge of track; otherwise there is only
local stratum water of little importance.
Figure 1. Plan view tunnel situation (Quick et al. 2001).
459
Figure 2. Geotechnical longitudinal section Old Tunnel Mainz.
Figure 4. Geological cross section.
4
Figure 3. German core centre technique OldTunnel Mainz
(Maidl 2004).
3
3.1
OLD TUNNEL MAINZ

Construction
The Old Tunnel Mainz was erected in the years 1881

until 1884 as one continuous double track tunnel. The
tunnel has an horseshoe shape and the lining consist
of sandstone with a thickness of approx. 0.9 m. The
German core center tunneling technique was used to
built the Old Tunnel Mainz with an overall length of
1200 m. The German core center technique (fig. 3)
uses partial drivings, mostly sidewall drivings. Due to
the bad condition of the tunnel, especially the masonry
and to improve the smoke venting system, a 300 m
long and up to 26 m deep open cut was built in the
early 30ies of the last century, which divides the tunnel
nowadays into the Tunnel Mainz Central Station and
the Tunnel Mainz south (fig. 1/2).
The Old Tunnel Mainz consists of the following
structures:
Tunnel Central Station: 663 m
Open cut: 300 m
Tunnel Mainz South: 246 m
CONSTRUCTION OF THE NEW TUNNEL
4.1 Construction techniques

Parallel to the existing Old Tunnel Mainz, the New
Tunnel Mainz was built in the years 1998 to 2000.
The new 1250 m long double track railway tunnel
with a low overburden of 10 m to 23 m runs under
buildings including a hotel with basements up to
10 m under ground level. Moreover, there are old
(Roman) underground hollow spaces (gallery systems) to be undercrossed.
The clearance between the Old Tunnel Mainz and
the new one varies between 4 m and 50 m (fig. 4).
Regarding ground conditions, existing settlementsensitive structures and the possible influence on the
Old Tunnel Mainz the excavation of the New Tunnel
Mainz had to be carried out only with little deformation. Hence, an universal shotcrete tunnelling method
with side wall drifts followed by the excavation of
the calotte and core/bench was chosen as construction
method (fig. 5).
The distance between the side wall faces and the
final lining (ring closure) was limited to less than
100 m and in particularly sensitive parts to 50 m.
The distance between the calotte face and the ring
closure of the preliminary support was restricted to
30 m. Apart from the usual measurements in tunnelling additional securing measures were applied in
460
Figure 6. Principle of the alpha-method (Quick et al. 2001).
Figure 5. Undercrossing of a hotel; driving concept and

securing measures (Steiger et al. 1999).
areas of settlement-sensitive structures. They are as

follows:
Horizontally injected steel pipe roof shelter
(strengthening of the longitudinal rock bearing
arch) (fig. 5). The roof shelter is placed in the upper
area of the face. The length of the drilling is 20.5 m.
The minimum overlapping to the next roof shelter is
3 m. The advantage of this technique is quite very
obvious; a widening of the roof area to place the
drillings is not necessary.
In order to prevent any dilatational effect above the
tunnel roof, 45 degree inclined, grouted PVC-fans
are installed in continuous distances of about 5 m.
Injections from the ground surface, pre-installed
injection systems under the foundation of buildings
as well as systematic face boltings within the tunnel
are applied additionally in order to minimize deformations. All these measures together in connection
with prior determined combinations of available
measures were the basis of a successful driving
with little deformations under settlement-sensitive
structures.
4.2
Calculation method
For proof of the stability and serviceability 2D- and

3D-numerical calculation were carried out by means of
the Finite-Element-Method with the program Abaqus
(figs 7, 8). Continuum elements were used for the soil,
where as beam elements for the lining. For the realistic
simulation of the soil an elastoplastic soil behaviour
was chosen (Quick et al. 2001). The modified
Figure 7. 2D-finite element mesh of the New Tunnel Mainz.
Drucker-Prager material law with cap was implemented. The yield surface of this elastoplastic model is
not constant in the principal stress space. It can expand
due to plastic straining. Furthermore it distinguishes
between different stiffness for loading, unloading resp.
reloading.
For the calculation of the preceding deformations as
well as to account for three dimensional arching effects
around the unsupported tunnel the alpha-method was
applied (fig. 6). The principle of this method is to
reduce the stiffness of the finite elements which are to
be removed in the next calculation step. The reduction
causes changes in the initial stress field and therefore leads to preceding deformation. In case of the
sidewall and calotte drivings the factor is set to 0.5.
This assumption which controls mainly the preceding
deformations was verified by measurements (fig. 11).
The 3D-finite element model was created by extruding the 2D-mesh. Under respect of the construction
procedure the length was chosen to 100 m (fig. 8).
4.3 Monitoring
Regarding the extraordinary situation to undercross
several settlement-sensitive structures with only low
461
Figure 8. 3D-finite elements mesh of the New Tunnel

Mainz.
Figure 10. Surface settlement at surface due to tunneling.
Figure 9. Measured surface settlements due to drivings

(TM 117).
Figure 11. Comparison of measurements and calculations
(roof displacements).
overburden in soft ground an extensive geotechnical

monitoring program had been carried out. At the surface the deformations due to tunnelling are measured
in close distances by levelling as well as deformation monitoring systems, working on the principle of
corresponding tubes. Figure 9 shows the measured surface settlements due to the driving at station TM 117.
The surface settlement adds up to 5 cm. The settlements measured at the ground surface along the tunnel
(fig. 10) were in most areas between 1.5 cm and 2.5 cm
in average; at the very beginning of the driving additional securing measures as described prior have
not been applied; the settlements at surface reached up
to unacceptable 11 cm.
Figure 10 shows the surface settlements under
respect of the different drivings (sidewall drift, excavation of crown etc.) at station TM 117. This station
is close to the portal north (fig. 1). Most of the measured surface settlements are related to excavation of
the sidewall drifts and the crown, while only a smaller
amount of settlements arise from the bench/invert
excavation.
Figure 11 shows the comparison between the measured roof displacements of the tunnel and the calculated roof displacements at station TM 117. The results
of the 2D-calculation show a good correspondence
with measurements regarding the preceding displacement as well as the displacements due to the excavation
of the sidewall drifts and the crown. However the calculated heave of the roof due to the excavation of the
bench/invert does neither match the measurement nor
the expected ground behaviour (fig. 11). For such soft
soil conditions it is therefore recommended to increase
the stiffness of the finite elements below the tunnel.
REHABILITATION OF THE OLD TUNNEL
5.1 Construction technique

The rehabilitation and enlargement of the Old Tunnel Mainz is going to be done under respect of the
462
experiences of the New Tunnel Mainz. The partial

loose ground (old back filling) around the Old Tunnel is improved by injections. In the first step the old
filling around the tunnel is injected. Voids will be
filled, improved by injections (bulk filling). For this
10 drillings up to a length of 2.5 m in a longitudinal
distance of 1.5 m are carried out (fig. 12).
In the second step up to 8.5 m long injection
drillings around the tunnel to activate a support ring
are drilled. These two ground improvement steps are
done under protection of the existing Old Tunnel. Subsequently the calotte driving in shotcrete method with
an additional forepiles takes places. For the stability
of the preliminary lining of shotcrete with a thickness of 0.30 m a widening of the calotte footing was
established. In the final construction step the invert is
excavated and the preliminary ring closure is achieved.
The distance between the face of the calotte driving and the preliminary ring closure with shotcrete is
limited to less than 12 m.
5.2
Calculation method
For the design and the proof of the serviceability

of the tunnel and the mentioned structures 2D- and
3D-numerical calculation are carried out with the program Plaxis. The numerical calculations regard all
construction phases as well as the former excavation
of the Old Tunnel Mainz and the New Tunnel Mainz.
In order to create realistic results the material law of
Hardening Soil with a yield surface, which is not fixed
in the principal stress space is used. The yield surface
expands due to plastic straining. In addition the material law can distinguish between different stiffness for
loading and unloading resp. re-loading.
To account for the three dimensional arching effect
of the unsupported enlargement of the Old Tunnel
Mainz the beta-method is applied under respect of the
used calculation program. The principle of this method
is described in 3 steps (fig. 13):
1. Generation of the initial stress field .
2. De-activation (excavation) of the tunnel clusters
without activation of the tunnel lining and generation of (1-beta) reduced forces, which can be
done by a reduction of the ultimate level of the full
calculation step.
3. Activation of the tunnel lining.
Figure 12. Tunnelling technique Old Tunnel Mainz.
Figure 13. Principle of the beta-method.
The beta-value was obtained by an iterative back

analysis of the New Tunnel Mainz. The predicted surface settlements of the 2D calculation for the ground
surface amount to approx. 2 cm (fig. 14).
5.3 Monitoring
The rehabilitation of the Old Tunnel Mainz is accompanied by an extensive monitoring program within the
Figure 14. Calculation results Old Tunnel Mainz.
463
tunnel and on ground surface. For a quick interpretation of the data three levels of settlement limits for
different areas under respect of the overburden and
the existing structure were defined. On the basis of
pre-defined measures such as an additional temporary shotcrete invert settlements can be slowed down
and reduced to guarantee the stability of the existing
structures.
6
CONCLUSION
The presented tunnels in the inner city of Mainz in

soft soil conditions show a variety of different measures in the ground for the drivings in order to meet
the requirements of the serviceability of existing buildings. 2D or 3D calculations were carried out to predict
the displacements. It is shown that the evaluation of
the input parameters and the calculation methods is
complicated but decisive for the calculated results.
Hence, every tunnelling project must be accompanied
by an intelligent monitoring program to observe the
impact on the environment due to the drivings and
to ensure the stability and serviceability of the tunnel and neighboring structures as well as to generate
data for possible back-analysis in order to improve the
calculated results and to verify the assumptions. The
paper shows also numerical approaches with different
material laws and calculation methods. These different
approaches can all lead to tolerable results, if methods are used properly and the input parameters are
evaluated appropriately.
excavation of a station of Toulouse subway line B

5TH international symposium, Amsterdam, Geotechnical Aspects of Underground Construction in Soft Ground.
June 2005
Kovacevic, N., Edmonds, H.E., Mair, R.J. & Higgins, K.G.
1996. Numerical modelling of the NATM and compensation grouting trials at Redcross Way Geotechnical
Aspects of Underground Construction in Soft Ground,
Rotterdam, 1996
Maidl, B. 2004. Handbuch des Tunnel- und Stollenbaus. 3.
Auflage. Verlag Glckauf GmbH, Essen
Quick, H., Michael, J., Arslan, U. 1999. Tunnelling for
German High Speed Railway Lines Proc. Civil and Environmental Engineering Conference New Frontiers and
Challenges, Bangkok, Vol. 2, part I, pp. II117128
Quick, H., Michael, J., Arslan, U. 2001. About the effect
of preliminary measures on ground movements due to
tunnelling Response of Building to Excavation-Induced
Ground Movements, London, 1718. July 2001
Quick, H., Meiner, S., Michael, J., Arslan, U. 2001.
Vergleich von Ergebnissen numerischer Berechnungen
mit in-situ Messungen am Beispiel eines Tunnelvortriebes STUVA-Tagung, Mnchen, 1922. November
2001, Studiengesellschaft
Quick, H., Michael, J., SchttnerV.,Arslan, U., Katzenbach, R.
2000. Tunnelling and Deep Excavation in Soft Ground
GeoEng2000, Melbourne, 1924. November 2000
Steiger, H., Theissen-Wenzel, C., Quick, H. 1999. Neuer
Mainzer Tunnel Behandlung geotechnischer Grenzflle
in der Planung und Ausfhrung Vortrge zum 6.
Darmstdter Geotechnik-Kolloquium, Darmstadt, 11.
Mrz 1999, Mitteilungen des Institutes und der Versuchsanstalt fr Geotechnik der Technischen Universitt
Darmstadt, Heft Nr. 44
REFERENCES
Emeriault, F., Bonnet-Eymard, T. & Kastner, R. 2005. Movements induced on existing masonry buildings by the
464
Supervision and protection of Shanghai Mass Rapid Line 4 shield tunneling

across the adjacent operating metro line
R.L. Wang & Y.M. Cai
Shanghai Metro Operation Co. Ltd., P.R. China
J.H. Liu
Shanghai Municipal Engineering Administration Bureau, P.R. China
ABSTRACT: Shanghai Mass Rapid Line 4 shield tunnels cross the operating Line 2 with 1.03 m distance
beneath the operating line 2, small angle and small turning radius. The tunneling-across was carried out without
ground improvement. This project took great challenge to the operating line. The influence of the tunnel crossing
was predicted and a scientific construction scheme and measures were arranged to guarantee safe operation of
Metro Line 2. The construction was performed based on the following principle strictly to push step by step
slowly, to turn equably with short step, to maintain stable pressure, and to improve the foundation with lower
pressure and small amount. Thanks to these measures, the shield tunnel crossed Metro Line 2 successfully.
Zhangyang RoadPudian Road tunnel of Metro

Line 4 was constructed with shield method. The shield
tunnel crossed under the operating Line 2 the operating line with a small angle and small turning radius.
Figure 1 shows you the # shape of the relation
between Metro Line 2 and 4. Following difficulties
were encountered in this project.
1. Distance. The new tunnel cross blow the existing
tunnel with minimum distance 1.03 m.
2. The tunnel at the crossing area is curved with small
turning radius. The radius of Line 4 tunnel is only
380 m. For the shield, to advance with a small
radius curve would influence the surroundings
more seriously than advancing along straight line.
3. Influence area. For Line 2, the shield advance will
directly influence 60 m at the up direction line and
94 m at the down direction line. Length of the
influenced range, in which the distance between
the upper and lower tunnel is less than the tunnel
diameter, is above 300 m.
4. No ground improvement. The construction point
was located in a place where several roads cross, so
that the ground traffic is very heavy. It is impossible
to conduct any ground improvement.
5. Bad geological condition. Line 2 tunnel is located
in the fourth layer stratum, which exhibits high
compressibility and obvious rheological behavior.
6. No relevant experience. Prior to this project, the
tunnel of Line 2 had ever crossed below Line 1
Figure 1. Plane and profile of the cross section.
when constructing, but in that case the ground

improvement was carried out. To sum up, tunnel
crossing without soil improvement is highly risky.
2
2.1
PROCESS ANALYSIS AND

COUNTERMEASURE OF THE SHIELD
TUNNELING
Process analysis
The influence of construction to metro operation was

predicted and the change of construction character was
465
Figure 2. Deformation of tunnel Line 2.
analyzed. The theoretical analysis shows: before the

shield-advance reach the centre line of Line 2 tunnel,
it would disturb the equilibrating stratum stress and
change the stress distribution. It would induce upward
and forward deformation to Line 2. When the shield
face passed through, the Line 2 tunnel would deform
continuously due to the reasons including excess pore
water pressure dispersing, consolidation settlement
and secondary consolidation settlement and operation
trains vibration. In order to control the deformation of
Line 2 tunnel, it is necessary to conduct synchronizing grouting and second grouting to Line 4 tunnel. The
grouting could give a pre-heaving to Line 2 and the further settlement could be smaller. If the operating tunnel
settlement approached the limitation, a re-grouting
would be needed. Meanwhile ground improvement
could be carried out between two tunnels to reduce
settlement of Line 2. Consequently, when the shield
crossed the Line 2 tunnel, longitudinal deflection in the
horizontal and vertical direction might be occurred.
It is composed with tunnel horizontal displacement,
vertical heave or settlement and radial convergence
deformation. It is shown in figure 2.
2.2
Main disturbing factors in the shield advance

process and its control measure
Indeed, various factors would cause Line 2 tunnel

deformation, the factors, however, could be reduced
to the following three categories:
1. During the shield advance process, the shield shell
and some projecting parts on the shield shell
would cause stratum loss. The stratum loss volume, marked as Ve1, is closely relevant with the
shield advance velocity. In order to reduce the stratum loss, we must limit the shield advance velocity
strictly and try our best to make it suitably slow
and constant. According to the theoretical analysis and the existing experience, the shield advance
velocity was set as v = 510 mm/min. Furthermore, we request the shield advance must pause
for 1030 min every 10 min advance. According to
the metro protection standard mentioned above and
the monitoring data from automatic equipments, we
would adjust and re-determine the pause time at any
moment.
2. Volume loss caused by curved shield advance is
marked as Ve2 . When the shield is advancing in
curve, in every 10 cm advance, an angle deviation, marked as , should occur at the shield axes.
Volume of Ve2 varies directly as the shield length
square and inversely as the curve radius. To reduce
the volume loss, articulated equipment of shield
machine must be used correctly, to reduce the
shield effective length. The deviation-correcting
degree per time must be limited strictly to ensure
Ve2 as small as possible. Theoretically speaking,
deviation-correcting at every ring advance with
1200 mm ring width is 17.6 mm. When it is divided
into 12 times, however, the deviation-correcting
volume per time is only less than 1.5 mm. That is
to say the small degree and multi-times deviationcorrecting would influence only 1/12 area as influenced by the original way. The disturbed area is
therefore considerably reduced and the volume loss
volume, Ve2, is also reduced consumedly.
3. Volume loss caused by unbalanced pressure at the
shield face, marked as Ve3 . Ve3 is the volume loss
caused by difference between shield face pressure
and stratum original static pressure P. During
the course that the shield advance is approaching
Line 2 tunnel, in order to reduce the metro tunnel settlement caused by shield tail grouting lack
and consolidation settlement, P should be a positive volume, namely the shield face pressure should
be 05% larger than stratum original static pressure. Because the Line 4 tunnel is in a small radius
curve in this section and it crosses the operating
Line 2 with a small angle, pressure on the shield
cutter head is unbalanced at left and right sides
during the shield crossing through process. And
sudden pressure change might occur during the process. To maintain the pressure at shield face stable
(especially when the segment is erecting, the jack
would be getting loose) and to make the soil be a
little uplifting is very beneficial to Line 2 tunnel
settlement control. Consequently, whether the construction is successful or not is depend on whether
the soil pressure at the shield face could be controlled and maintained at a suitable level. What
should be adjusted in the construction process is
the parameter mentioned above: the shield advance
velocity, v; pressure at shield face; and axes deviation occurred in every 10 cm advance, , (generally,
this index is defined as the predetermined volume).
General effectiveness caused by these three factors,
when reflected in the Line 2 tunnel horizontal displacement, vertical uplift or settlement and tunnel
radial convergence, is v h and the tunnel radial
convergence deformation.
According to the metro tunnel deformation specification and metro safety operation requirement, for
the operating Line 2 tunnel, the maximum tolerance
of the longitudinal deflection in the horizontal and
vertical level planes are determined as 5 mm; and
466
pressure and small amount. The technology keys are

decomposed into the following points:
the tunnel radial convergence deformation tolerance

is strictly controlled within 20 mm.
2.3
Real time information construction
During the shield advance process, the vertical deflection of the Line 2 tunnel, both in vertical level
horizontal level, marked as v, and in the vertical
plane, marked as h must be controlled strictly. Whenever the shield advances a step, the change rate of v
and h must be analyzed and the construction parameters must be adjusted within 5 mm tolerance, so
that to control the tunnel deformation. In the shield
advance process, high-automatic and high accurate
monitoring system was adopted to supervise Line 2
tunnel status lively. High-precise automatic electrical level supervision system was adopted to measure
the tunnel longitudinal settlement and transverse subsidence difference between two tracks; and Bassett
convergence system adopted to supervise the tunnel
convergence status. After the train service finished, we
adopt additional manual supervision to monitor settlement, displacement, and convergence status of the
Line 2 tunnel (to set 3 settlement monitoring points
at each profile; 1 on the track bed and the other 2 on
the segment the monitoring profiles are set every 2 m
along the tunnel.)
2.4 Analysis on grouting and foundation
improvement
After the shield cross over the Line 2 tunnel centre
line, whether the interspace at the shield tail could be
refilled timely and suitably is a key factors that cause
tunnel longitudinal deformation. In this process, the
shield jack stress on the tunnel segment would cause an
opposite longitudinal deflection, but control volume of
h must be less than 5 mm. After the shield passed,
the consolidation settlement of the layer beneath the
tunnel is also an unneglectable factor. Therefore, after
the shield passed below the Line 2 tunnel, several
timely grouting with low pressure and small amount
is very important and directly related to the success of
Line 2 tunnel settlement control.
3 TECHNOLOGY KEYS OF THE SHIELD
ADVANCE AND PRE-CONTROL OF MAIN
CONSTRUCTION PARAMETER
Based on the deep analyses and clear understanding on
the construction difficulties, the operator conducted a
45 kph speed limit in the relevant section of Metro
Line 2.
We comply with the following principle carefully
and strictly in the construction to push step by step
slowly, to turn equably with short step, to maintain stable pressure, and to improve the foundation with lower
1. Settlement or uplift control volume of the Line 2

tunnel is 5 mm. When the uplift volume reaches
50% and settlement, 30%, of the control volume,
the monitoring system will alarm. That makes the
tunnel always in a slight deformed condition.
2. The shield advance slowly and constantly;
the shield advance speed is maintained at
510 mm/min. And the shield advance must pause
for 1030 min every 10 min advance. The pause
time length depends on whether the Line 2 tunnel
has uplift fall back to the protection standard.
3. The shield face soil pressure should be adjusted to
a correct volume so that it would cause minimum
influence onto the Line 2 tunnel. To maintain a stable face soil pressure is extremely important. The
shield face pressure should be 100105% of stratum original static pressure, however, the volume
is flexible and should be adjusted based on the live
monitoring data. When the segments are erecting,
there should be effective precaution to prevent the
face soil pressure from descending; and the erecting
time should be as short as possible.
4. Shield Advance Deviation-correcting. The
deviation-correcting degree per time is the smaller
the better. The deviation-correcting is required
to taken every 10 cm advance. And the articulated equipment of shield machine must be used
correctly, to reduce the shield effective length.
5. Synchronizing grouting pressure should be less
than 0.4 mpa or even lesser. Grouting amount is
1.11.8 times of the interspace cubage. Detailed
grouting amount depends on the monitored Line 2
tunnel data.
6. Second liquid grouting would be adopted in the
second grouting. The grouting conduct principle is:
low pressure, multi-times, moderate and timely.The
pressure should be maintained at 0.2 mpa. Grouting
amount per time per ring is 50100 liter. We request
the grouting should be conducted every other ring,
to reduce its influence to Line 2 above. Generally,
we request the grouting only be conducted when
the train service finishes.
7. Foundation improvement is necessary to prevent
the Line 2 tunnel from settling continuously. In this
process, all relevant parameters such as, grouting
pressure, grouting amount, grouting time, grouting
pipe withdrawing velocity, and grout mix proportion and consistency, should be limited carefully to
reduce the harmful influence to the tunnel above as
much as possible.
8. Set automatic and high-precise electrical level system along the Line 2 tunnel. The electrical level is
set to acquire data every 5 minutes and deliver the
data to shield operation control centre. The operation centre would adjust construction parameter and
467
Tunnel Deformation Control
Beyond the Tolerance
Construction Parameter Adjustment
Live Automatic Monitoring
Compare the Collected Data with the

Tolerance
Figure 4. The Line 2 up line tunnel settlement/uplift data

during the first tunneling cross process.
N
Within the Tolerance
Construction Goes On
Figure 3. Pre-control system sketch.
guide shield construction based on these data. The

construction technology keys and quantified construction parameter are together formed an effective
and reliable pre-control system, which is shown in
figure 3.
4.1
Figure 5. The Line 2 down line tunnel settlement/uplift data

during the first tunneling cross process.
SUPERVISION ON THE SHIELD

CONSTRUCTION AND THE OPERATING
TUNNEL DEFORMATION ANALYSES
4.2 Process description and site monitor of the first

tunneling cross
The shield machine is 8.625 m long and the outer diameter

= 6.34 m. The segments are 350 mm thick and
single segment longitudinal length is 1.2 m. The whole
construction process is as follows:
1. The first tunneling across process of the shield.
Shield at Line 4 down line tunnels across Line 2
up line first, and then tunnels cross Line 2 down
line.
2. After the shield machine arrived Zhangyang Road
Station, soil between the tunnels at the crossed
section was improved to ensure the metro Line 2
operation safety.
3. The second tunneling cross process of the shield.
The shield turned back to advance in the Line 4
down line. It cross Line 2 down line first, and then
tunnels cross Line 2 up line.
4. After the tunneling finished, soil of this section was
improved. Four tunnels at the cross through section
formed a # shape with small angle on a projection
plane. The project therefore influenced very large
range of Line 2. Length of the range, in which the
distance between the upper and lower tunnels is less
than the tunnel diameter, is above 300 m.
The first shield tunneling cross process was started

from Pudian Road Station. The first tunneling cross
process lasted for 15 days.
At the very beginning of the shield advance, the preset construction parameter makes the Line 2 tunnel to
maintain a slight uplifting trend.
When the shield cutterhead was approaching below
the Line 2 up line tunnel projection, the shield advance
velocity and face soil pressure were adjusted, but
the Line 2 tunnel still maintained uplifting trend and
the uplift volume was up to 0.20.5 mm. During the
shield advance process, soil at different points had different stress, thus caused different influence in the Line
2 tunnel. When deviation-correcting conducted, soil at
head and left of the shield machine was stressed but
stress of soil at head and right was released; meanwhile, at the shield tail, soil at the right side was
stressed but stress of soil at left side was released.
When the shield advanced just below the axis of
Line 2 tunnel (the shield was only 1.44 m away below
Line 2 tunnel then), the face pressure onto soil had been
adjusted speedily from the previous 0.230.27 mpa to
0.166 mpa, much smaller than the previous pressure.
468
Figure 6. Settlement/uplift data of the closest point from the

Line 2 up line tunnel during the first tunneling cross process.
Figure 8. The Line 2 down line tunnel settlement/uplift data

during the second tunneling cross process.
Figure 7. Settlement/uplift data of the closest point from

the Line 2 down line tunnel during the first tunneling cross
process.

during the second tunneling cross process.
At that time, the up line of Line 2 tunnel had uplifted

for 2.53.1 mm.
When the shield cutterhead was escaping from
the projection of Line 2 tunnel, because of incorrect
and unsuitable deviation-correcting and synchronized
grouting, instantaneous uplift volume of Line 2 tunnel was up to 3.3 mm. Figure 4 and figure 5 show
the settlement and uplift condition during shield cross
through Line 2 up line tunnel and down line tunnel
respectively.
Figure 6 and Figure 7 show the settlement and uplift
condition of the closest settlement point during shield
cross through Line 2 up line tunnel.
escaped from Line 2 projection completely. The second shield tunneling cross process totally lasted for
23 days.
The shield crosses below the Line 2 down line tunnel first with the minimum vertical distance 1.03 m.
Its influence on the Line 2 tunnel is obvious, although
we have made very strict limit on the construction
parameter. By July 1, the Line 2 down line tunnel has
uplifted up to 2.5 mm. Because the up line tunnel of
Line 4 is very closed to Line 2 tunnel, the Line 2 tunnel assumed obvious dynamic settlement during cross
through process. It can hardly be controlled except by
synchronized grouting and second grouting method.
Figure 8 to Figure 11 shows detailed condition.
4.3 Process description and site monitor of the

second tunneling cross
4.4 Grouting and foundation improvement issues
The second shield tunneling cross process was started

from Zhangyang Road Station. By June 24, 2003,
when the shield of Line 4 construction arrived 2 m
away from the projection line of the Line 2 down line
tunnel, the shield advance had caused settlement and
uplift at Line 2. On June 30, the shield cutterhead
entered the projection line of Line 2 down line tunnel, then it crossed over below the Line 2 down line
and up line one after another. By July 22, the shield
In the shield advance process, in order to control

the Line 2 tunnel settlement after shield advance,
unreactive-grout-synchronized grouting was conducted, so that the grout would refill interspace at
the shield tail. Meanwhile, synchronized grouting and
second grouting was conducted at the cross section
where dynamic settlement occurred. Double liquid
grout was used in second grouting to improve the
stratums mechanical properties. 16 grouting holes are
469
repeating foundation improvement in the recent year,

the Line 2 tunnel settlement situation is now well in
hand.
5
CONCLUSION
After the elaborate construction and strict site supervision and protection, the Line 2 tunnel settlement
situation is now well in hand and the safe metro operation during the project are also realized. Valuable
experience, which could be reference for the future
construction, was accumulated in this case.
Figure 10. The Line 2 down line tunnel settlement/uplift
data at every time during the second tunneling cross process.

at every time during the second tunneling cross process.
1. When the shield advancing in curve with small

radius, because of different stress condition at different point of the shield, it would cause various
influence to the upper metro tunnel.
2. The shield advancing velocity is an important factor that causes upper tunnel uplift. The influence,
however, could be reduced by suitable construction
parameter adjustment.
3. Shield deviation-correction is another important
factor that causes upper tunnel uplift. The uplift
or settlement condition is dependent on relative
position of the two tunnels.
4. Shield tail grouting would influence quite a large
range, up to 1015 m. It is necessary to maintain
low pressure, small amount and correct grout in
the grouting process, to reduce the influence to the
surroundings.
5. Dynamic settlement caused by train running is very
dangerous. Synchronizing grouting and second
grouting could be adopted to correct that. Foundation improvement and double liquid grouting are
important method to control tunnel settlement.
6. The principles to push step by step slowly, to turn
equably with short step, to maintain stable pressure,
and to improve the foundation with lower pressure
and small mount are proved correct technical keys.
7. Strict high-precise automatic monitoring and information based construction is one of the most
important technical support.
Figure 12. Schema of the Foundation Improvement.
ACKNOWLEDGEMENTS
reserved at every segment in the cross section, so that
it would be easy to conduct stratum grouting improvement after the shield escaped from the section. The
arrangement of the foundation improvement grouting
point is shown in figure 12. After the shield passed, by
I would like to extend grateful appreciation to

Mr. BaiTinghui, Mr. Ge Shiping, Mr. Shen Chengming,
Mr. Sun Lianyuan and Mr. Ni Chengyu for their
assistance during composing this thesis.
470
Kowloon Southern Link TBM crossing over MTR Tsuen Wan Line
tunnels in HKSAR
K.K.W. Wong, N.W.H. Ng & L.P.P. Leung
Kowloon-Canton Railway Corporation, HKSAR
Y. Chan
MTR Corporation Limited, HKSAR
ABSTRACT: The Kowloon Southern Link (KSL) is a 3.8 km underground railway project in a busy urban area
with complex geological and alignment constraints.A key challenge for the project is to construct a critical section
of the tunnels above the existing MTR tunnels and below a heavily trafficked road junction, which is termed as
the MTR Crossing. Cut and cover method was originally envisaged to construct the MTR Crossing. In view of
the complex ground conditions and other physical site constraints, the contractor proposed to extend the bored
tunnels using a tunnel boring machine (TBM) across the existing MTR tunnels as an alternative. Adopting the
TBM method for the MTR crossing introduced specific challenges which called for comprehensive engineering
studies, extensive ground treatment, tailored protective measure for the MTR tunnels, as well as sophisticated
instrumentation in order to ensure that the railway operations and the public safety are not compromised. This
paper is to discuss the design development of the scheme and the methodology of the MTR Crossing.
INTRODUCTION
The Kowloon Southern Link (KSL), running 3.8 km

underground, will link West Rail from its current
terminus at Nam Cheong Station to the East Rails terminus at East Tsim Sha Tsui Station, with a new West
Kowloon Station in between. By joining East Rail with
West Rail, it will provide commuters with a convenient
interchange between these two railway corridors.
The KSL construction works are basically divided
into two sections, northern section and southern section. The northern section covers a twin 2.1km running tunnels between Nam Cheong Station and West
Kowloon Station, mainly on reclaimed land where the
conventional cut and cover construction is the prevailing method. The 1.7 km southern section comprises
a new station and twin 1.1 km TBM bored tunnels
located in a busy tourist and commercial area between
West Kowloon Station and East Tsim Sha Tsui Station. The project includes the crossing perpendicularly
above the existing Mass Transit Railway (MTR) Tsuen
Wan Line (TWL) tunnels which run in the north-south
direction underneath the heavily trafficked Nathan
Road (Figure 1).
This paper focuses on the TBM crossing over the
MTR TWL tunnels.
Figure 1. The KSL Alignment.
SITE CONDITIONS AND REGIONAL

GEOLOGY
The solid geology in the Kowloon peninsula and along

the KSL tunnel route and at the MTR crossing comprises intrusive igneous bedrock belonging to the
Kowloon Granite Formation. Salisbury Road is 4 m
above sea level and constructed over old reclaimed fill
underlain by superficial deposits. The deposits overlie
471
Figure 2. Geological sections.
a relatively thin layer of completely decomposed granite (CDG) above bedrock. Despite close proximity of
the harbour the ground water table in Salisbury Road
was typically 2.5 m below ground level with minor
influence from tidal fluctuations.
The geological interpretation indicated that the
MTR down track tunnel lies within the granite
bedrock. For the MTR up track tunnel, as-built record
indicates the bedrock and completely decomposed
granite interface to be at the axis of the MTR up
track tunnel. However, based upon additional geotechnical information and borehole results, this interface
was interpreted to be higher above the tunnel axis
(Figure 2).
3 TUNNEL DESIGN
3.1
Original design based on cut and cover
The original scheme design of the KSL tunnels

along Salisbury Road was based on the conventional cut-and-cover construction method. KCRC, the
project proponent, submitted its original proposal
based upon the original scheme design to the HKSAR
Government.
However, MTR raised an objection to the proposed
scheme for the reason that the proposed cut-and-cover
method at the MTR crossing would give rise to the
risk of floatation due to removal of the overburden
above the MTR tunnels at the crossing. Subsequently,
an alternative micro-tunnelling technique was further
proposed for the MTR crossing.
The proposed micro-tunnelling method required
hand mining in soft ground within a pre-installed horizontal pipe piled enclosure around the perimeter of the
tunnel structure. It was contended that this alternative
method would ameliorate the risk of floatation of the
existing MTR tunnels compared to the cut-and-cover
method. As such, MTR agreed to allow authorization
of the scheme by the HKSAR Government based on

the micro-tunnelling method.
Despite MTRs no objection to the use of the microtunnelling method from railway protection point of
view, there were drawbacks of this method. It is recognized that hand mining in soft ground, in particular in
marine deposits and below groundwater table, is a high
risk activity. The risks of hand mining include the indirect support system, instability of mined face, potential
for seepage and piping through the gaps between horizontal pipe piles, and the likelihood of large ground
movements resulting in a risk of damaging existing
utilities. To mitigate these risks, substantial ground
treatment would be required for the micro-tunnelling
construction at the MTR crossing. Furthermore, excavation by hand mining with staged installation of
temporary supports and construction of the permanent
tunnel structure would be time consuming.
During the KSL tendering stage, some of the tenderers proposed the use of TBM method for the MTR
crossing. Due to the limited time for the tendering,
none of the tenderers was able to obtain MTRs agreement on adopting the TBM method. As such, the TBM
crossing as a potential alternative remained uncertain
at the time of the tendering.
3.2 Selling the TBM crossing alternative
Immediately after the contract award, a value engineering workshop was held among KCRC, the contractor
and its designer. Consensus was reached that the TBM
crossing alternative should be seriously considered.
The contractors designer undertook a detailed engineering study of the TBM crossing over the MTR
TWL tunnels. The study included a two-dimensional
finite element numerical modeling analysis using the
software PLAXIS Version 8.2 and a three-dimensional
analysis using the software FLAC 3D Version 3.0.
In both the 2-D and 3-D analyses, the geotechnical
parameters adopted were based on the geology determined from existing site investigation information and
the MTR construction records. Due to the close separation of the two KSL tunnel bores, ground treatment
using jet grouting was recommended by KCRC and
its contractor (Figure 3). The analyses showed that
the movements and distortions of the MTR tunnel
lining due to the KCRCs TBM crossing would be
within the required structural tolerance. Furthermore,
a parametric study was carried out with varying jet
grout strengths between 0.5 MPa and 1.5 MPa and
it indicated that the impact on the MTR tunnel lining movement would be satisfactory and the jet grout
strength was not a critical factor.
Despite the favorable results of the design assessment, MTR expressed a specific concern regarding
the presence of boulders above the crown of the MTR
tunnels and below the invert level of the KSL tunnels. MTR considered that there was a risk of the TBM
472
Figure 4. KSL Alignment constraints.

Figure 3. Layout of Jet Grout Block.
dislodging a boulder which may be connected to the

MTR tunnel lining, resulting in damage to the MTR
tunnel lining. In order to mitigate this risk, it was proposed to drill closely spaced horizontal pipe piles to
physically isolate the MTR tunnels from the material
through which the TBM would bore. Furthermore, jet
grouting the entire area above the MTR tunnels was
proposed to form a consolidated block of soil and grout
matrix. On this basis, MTR agreed in principle to the
TBM method of crossing the MTR tunnels.
However, the alternative TBM proposal still needed
to undergo the statutory process of the amendment to
the authorized scheme, which also involved consultation with relevant Government authorities, stakeholders, as well as local communities. The merits of the
TBM method compared to the original scheme of cut
and cover tunnels were obvious from the public and
social benefit points of view, apart from the construction benefits. The TBM method minimizes disturbance
to the traffic and existing utilities with much less environmental impacts in terms of noise, dust and visual
impact, in particular in such a busy commercial and
tourist area where there are nearby prestigious hotels,
shopping malls and distinguished cultural facilities.As
such, the alternative proposal for adopting the TBM
method was well received by all the relevant parties
without any objection.
4 TECHNICAL CHALLENGES FOR TBM

CROSSING
KCRC and its contractor needed to revise the design
details of both the permanent works and temporary
works to cater for the revised TBM crossing method.
This included addressing the vertical alignment constraints imposed by the existing MTR tunnels, the
connection into the existing KCRC overrun tracks of
the ETS Station, as well as the ground conditions for
TBM tunnelling (Figure 4).
In order to maintain the minimum 1.5 m clearance

of any KSL works from the extrados of the MTR
tunnels, the depth of the ground cover to the TBM
at the MTR crossing was as shallow as only 6.8 m
which was less than the diameter of the TBM (viz.
8.05 m). Given the geology at the crossing with boulders present at the tunnel invert and marine deposits
within the tunnel horizon and above the tunnel crown,
the shallow cover is considered a challenge to the TBM
tunnelling in controlling the ground settlements on one
hand and minimising the risk of ground heave or slurry
leakage on the other. There were existing critical utilities above the crossing. The consequence of damaging
these utilities would be very serious. Also, any significant disturbance to the surface such as excessive
ground movements or slurry leakage would inevitably
cause disruption to the traffic at this busy road junction. These kinds of incidents would have major impact
on the road users and KCRCs corporate image and
therefore measures were taken to mitigate the risks of
their occurrence.
Due to the above vertical alignment constraint, the
structural clearance between the KSL tunnels and the
MTR tunnels was only 2 m. The presence of boulders
at the KSL tunnel invert had been a cause for concern on potential damage to the MTR tunnel lining.
The MTR TWL carries more than a million passengers daily. Apart from the public safety which is of
paramount importance, any disruption to the railway
services would have substantial impact to the public
and is unacceptable to MTR and Government authorities. Careful planning, control and monitoring of the
TBM drives crossing above the MTR tunnels were
deemed to be critical and essential in order to ensure
that there would be no adverse impact on the MTR
tunnels. These called for the need of the development
of wide range of additional protective measures and
geotechnical instrumentation.
The existing overrun tracks of the ETS station to the
east of the MTR crossing had been operational prior to
the KSL construction. The alignments of these overrun tracks had been based on the original cut and cover
473
the marine deposits in the area of the MTR crossing

were carried out in order to obtain sufficient data for
calculating the air bubble pressure to be adopted by
the TBM when crossing over the MTR tunnels.
6 ADDITIONAL PROTECTION TO MTR
TUNNELS
6.1 Ground treatment
There are multiple objectives of the ground treatment
above the MTR tunnels as follows:
To stabilize the ground whilst the TBM passes
above the MTR TWL tunnels and beneath sensitive
utilities,
To provide water cut-off and prevent lowering of
ground water and loss of soil when the TBM breaks
into the reception shaft, and
To cater for TBM low ground cover and narrow
separation
Figure 5. The mixed ground TBM.
tunnel scheme of the KSL with a very narrow separation between the up track and down track. he structural
clearance between the two KSL tunnel bores at the
MTR crossing was only 900 mm. This extremely close
separation of the tunnels gives rise to a technical challenge for the TBM tunnelling, in particular on how to
avoid causing excessive movements to the first completed KSL tunnel during the adjacent second TBM
drive.
5 TUNNEL BORING MACHINE
A purpose-built mixed ground TBM, using mixshield/
slurry technology, was used to excavate the tunnels
through both rock and soil deposits (Figure 5). The
8.05 m diameter Herrenknecht TBM, named Little
Dragon Girl (Xiaolognu), was designed to suit the
prevailing ground and groundwater conditions and so
limited the potential for face instability and ground
settlement when excavating in soft and mixed ground
conditions. The TBM cutter head was also equipped
with heavy-duty cutting tools for excavation in competent rock. A hydraulically powered jaw type rock
crusher was installed in front of the suction grill behind
the cutter head and was capable of crushing blocks
such as boulders up to 500 mm in size.
The operating principle of the mixshield TBM is to
utilise the pressurized bentonite slurry at the confined
tunnel face to withstand the hydrostatic pressure and
ground pressure in all instances. This system enables
the tunnel face stability to be maintained at all times,
even in a mixed soil/rock interface.
The bentonite suspension for the tunnel face support
is pressurised by an air cushion or air bubble. Little
Dragon Girl was equipped with a compressed air regulation unit which accurately controlled the required
air bubble pressure with a sensitivity of +0.05 bar.
Pressure levels were calculated for various sections of
the TBM tunnel drives according to the different geological and groundwater conditions. Extensive tests on
Ground treatment in the form of 1.5 m diameter

jet grout columns were adopted using cement based
grout (including microfine cements) and silicates. For
quality control purpose the unconfined compressive
strength of post-drilled cores was set at 1 MPa minimum at 28 days. Jet grout columns generally terminate
at 1.5 m from the crown of the MTR tunnels and 2.5 m
above the crown of the future KSL tunnels. The total
number of jet grout columns designed is 690.
The jet grouting employed a double fluid jet grouting system utilizing a swivel head that directs two
different fluids of grout and air through concentric
double tube rods. Grout and air exit horizontally
through the two concentric nozzles attached to the bottom of the grout string. The air and grout are injected
at high pressures of 6 bars and 200 bars respectively.
The air and grout injection breaks up and mixes with
the soil matrix resulting in a grout and soil slurry. The
diameter of the grout column is controlled by the rate
of rotation and withdrawal of the string. In general,
a rotation rate of 6 to 7 rpm and withdrawal rate of
1820 min/m were adopted to suit the soil stratum. In
the cohesive clay layer, precutting by water jet and
grouting was carried out at a slower drill rod withdrawal rate of 2022 min/m. Excess sludge and slurry
is removed through the annulus between the boreholes
and the drill rods by air lifting effect. The completed
column is formed as a hardened cylindrical body of
soil cement mixture.
Working space and access were severely restricted
due to the requirement to maintain traffic flow through
a busy junction as well as numerous underground utilities (Figure 6). It was necessary to divert the traffic
in multiple stages to facilitate jet grout column construction within the footprint of the available traffic
management schemes. A combination of vertical and
474
Figure 6. Jetgrouting in progress.
Figure 7. Horizontal pipe piles installed.
raking holes were constructed, depending on the traffic

and utility constraints.
Quality control testing was carried out to prove the
effectiveness of the ground treatment. Core samples
were drilled and collected from 5% of the columns for
unconfined compressive strength testing. Permeability
tests were also carried out in vertical drill holes from
ground surface and water inflow measurements were
made in horizontal probe holes from the TBM retrieval
shaft to verify the water-tightness of the completed jet
grout block. The results proved to be satisfactory.
6.2
Horizontal pipe piles
Horizontal pipe piles were installed through the jet

grouted zone above the MTR TWL tunnels to provide
physical separation between the TBM and MTR TWL
tunnels, and to isolate any core stone boulders which
might be disturbed by the TBM bore.
Conventional Down-the-Hole (DTH) techniques
were adopted for the drilling operation using an
ODEX bit. The horizontal pipes were 273 mm diameter 6.3 mm thick pipe piles installed at 500 mm centre
to centre spacing. The total length of each pipe pile
is about 23 m in order to extend over both MTR
tunnels and 5 m beyond the centre line of MTR up
track tunnel. Structural clearance of the horizontal
pipe piles to the MTR tunnels varies from a maximum of 2.0 mm to a minimum of 1.397 m. Tight
structural clearance is a combined result of the vertical alignment constraints as described in section 4
above and the downward deviation of the horizontal
pipe piles. Drilling of isolated piles which infringed the
railway protection zone were permitted however this
was conditional on additional requirements for working during non-traffic hours, detailed survey checks
and comprehensive monitoring.
Controlling the alignment of the horizontal pipe
piles is critical to ensure the pipe piles do not deviate downwards within 1.2 m of the MTR tunnels or
upwards into the TBM bore. Alignment control was
achieved using a down-the-hole survey method. A

survey prism is inserted into the far end of the piles, followed by survey checks every 3 m as the survey prism
is withdrawn from the pipe.The survey check is carried
out at 5 intermediate stages for each pile as follows:
(1) before drilling and after the set-up of the drilling
mast; (2) at 4 m depth before arriving at the MTR down
track tunnel; (3) at 10m depth between the MTR tunnels; (4) at 16m depth before arriving at the MTR up
track tunnel; and (5) after reaching the final length
of 23 m.
Various contingency measures were developed in
the event of recording unacceptable upward or downward deviation. These included replacing the pipe piles
with bundled rebar within a bentonite cement mix and
down-sizing the pipe piles to 219 mm in diameter.
INSTRUMENTATION AND MONITORING
The instrumentation and monitoring for the MTR

tunnels was required to meet statutory and specific contract requirements. In addition, Alert, Action
and Alarm Response levels (AAA) for building and
structure movement, utilities, ground settlement and
groundwater change were developed.
It is a pre-requisite requirement for the works within
the MTR railway protection zone that the MTR railway
and its continuous operation must not be jeopardized
and be safely maintained during the TBM crossing. As
revealed by the numerical modelling analyses, some
movements of the MTR tunnel lining may be caused by
the construction of the TBM tunnels directly above. To
ensure the safe and continuous operation of the MTR
railway at all times, an extensive regime of instruments and monitoring programme was established
both inside and outside of the tunnels. The monitoring
programme was a large part of the mitigation measures
for the crossing to minimize the risks of damage to the
MTR tunnel lining and consequential interruption to
475
busiest operating railways with minimum clearance,

but no train speed restrictions or interruptions to the
services and with an excellent safety record.
The site conditions and the existing railways pre Automatic Deformation Monitoring System (ADMS):
sented an unusually onerous range of technical chalreal-time deformation monitoring of the tunnel linlenges. The success of the MTR Crossing has estabing by automatic total stations within the influence
lished a new benchmark for the TBM tunnelling
zone at the crossing
technique and with such recognition received would
Vibrating Wire Strain Gauges: real-time monitoring
no doubt lead to its further potential use in complex
of changes in the stress of tunnel lining
urban environment similar to those found in HKSAR.
Seismograph: real time monitoring of ground borne
The project has strongly demonstrated that, for parvibrations induced by the works
ticular spatial and operational constraints, the TBM
Tape extensometers: manual convergence survey of
tunnelling method can provide the best overall value
the tunnels
and performance in constructing new underground
Manual track level offset surveys
infrastructures.
Instruments and monitoring of the ground surrounding the MTR tunnels:
ACKNOWLEDGEMENT
Vibrating Wire piezometers: real-time monitoring
of changes in pore water pressure
The MTR Crossing required also the detailed planning
Manual and automated in-place vertical inclinomeand substantial cooperation and partnering interaction
ters: monitoring deformation of the ground and
between the two railway corporations. The proactive
cofferdam walls
involvement from MTR in the selection of an appro In-place horizontal inclinometers: installed inside
priate day for the crossing minimizing all possible
selected horizontal pipe piles above MTR tunnels
adverse impacts, the provision of stand-by team and
monitoring real time ground movement
engineering trains during the crossing nights reacting
Magnetic probe extensometer: manually monitorto incidents, if any, instantaneously, and the carrying
ing of deep ground settlement.
out advanced improvement works inside TWL tunnels
before the crossing, are all contributing factors to the
All monitoring data was uploaded directly to an
success of MTR crossing which was delivered safely
instrumentation database system called GEOMON
and without interruption to the train services.
which is a proprietor product from the specialist instruThe authors would also like to thank the contractor
mentation subcontactor. GEOMON also managed the
Link 200 JV and their design team Mott Meinhardt JV
data by automatically alerting designated recipients
for the invaluable input in developing and finalizing
of AAA response value exceedances via e-mails and
the MTR Crossing scheme.
SMS, and could also be viewed and downloaded
remotely via the Internet.
Upon exceeding the AAA levels, procedures outREFERENCES
lined in the Exceedance Plans were implemented
accordingly. During excavation for the first of the
HKSAR Works Bureau, Environmental Transport and Works
two KSL tunnels the TBM successfully crossed over
Bureau, Technical Circular No. 19/2002 Mass transit
the MTR tunnels. Insignificant movement and ground
Railway Protection, May 2002.
borne vibration was recorded in the MTR tunnels.
the railway services and include the following types of
instruments and method of monitoring. Instruments
and monitoring within the MTR tunnels:
CONCLUSION
The change from traditional cut-and-cover construction to TBM method enabled two new railway tunnels
to be constructed directly above one of the worlds
476
Application of pile underpinning technology on shield machine crossing

through pile foundations of road bridge
Q.W. Xu
Shanghai University of Science and Technology, Shanghai, P.R. China
Shanghai Jiaotong University, Shanghai, P.R. China
X.F. Ma
Z.Z. Ma
Shanghai Shentong Metro Co. Ltd., Shanghai, P.R. China
ABSTRACT: Shanghais subway construction has entered a new period of rapid development. A notable
characteristic of subway network construction is that the construction environment is increasingly complex.
Take Shanghai subway line 10 as an example, in the position of Shajinggang Bridge on Siping Road, the shield
machine will have to pass through the bridges pile foundation. Since the embedded depth of piles has already run
through the entire tunnel section, these piles have to be pulled out or cut off. But, as one of the main arteries of
Shanghai city, Si Ping Road has big traffic flow, and it does not allow the traffic volume to be affected during the
construction period. Whats more, there are intensive residential quarters around the bridge and the construction
space is limited. Therefore, this paper will mainly introduce how the pile underpinning technology can be used
in the subway construction of Shanghai area for the first time.
INTRODUCTION
M10 Line
Huangpujiang River
R1 Line
Along with the continuous construction of metro lines

in cities and unceasing improvement of metro network,
there are more and more cases of new tunnel lines
passing through old nearby buildings [1][3] .
In the past 20 years, Shanghai has been experiencing
the unprecedented climax of city infrastructure construction. There are total 17 lines planed in the urban
rail transit network, of which 11 lines are subway with
length of 385 km, presently, 5 lines have been completed, 5 lines are being constructed. By 2010, total 10
lines will be put into use with length of 250 km.
A notable feature of the new period of metro
construction is that the construction environment is
increasingly complex, that is, examples of intercrossing between new line and old line, crossing through
various existing structures etc., are becoming more
and more. Take Shanghai subway line 10 as example, it
spans from Hongqiao Airport to New Jiangwan Town
with total length of 28.8 km, which forms a convenient
channel between the north and the west part of Shanghai city. During its construction process, the main pile
foundations that impede the advance of shield machine
are located at Shajinggang Bridge, Zoumatang Bridge,
M7 Line
M3 Line
Huangxing road
Zou Matang
Sha Jinggang
M4 Line
Suzhou River
R2 Line
Chengjia Bridge
R4 Line
M7 Line
Figure 1. The general picture of subway Line 10.
Chenjiaqiao Bridge and Huangxin Road Bridge, as

shown in Figure 1.
Regarding to the situation of new line crossing
through existing structures, generally speaking, the
line should be chosen to avoid existing buildings in
the stage of planning. In fact, there are various practical situations often make it difficult to avoid such
477
1584
0
1559
0
3210
0
2230
=2453
0
1553
0
2150
310
2150
2730
11
3210
0
2230
=2453
0
3000 1900
1553
0
1584
0
1559
0
11
3000 19
00
15
310
structures. As to the examples of tunnel crossing

the bridge foundation mentioned in this paper, when
the line can not avoid existing pile foundation, the
general construction method is: construct temporary
alternative bridge change traffic course, remove the
old bridge, extract pile drive the shield machine
forward construct a new bridge restore the original traffic demolish the temporary bridge.
If follow this method, it results in long construction period, high construction cost, and great impact
on the society. But if we adopt the pile underpinning technology, that is, based on the premise of
keeping the existing structure can be worked normally, while a series of construction technology are
used to extract or truncate piles, and finally reach
the purpose of driving shield machine forward. This
can not only shorten construction period, reduce construction cost, minimize its impact on the community,
but also can promote Shanghais subway construction
level.
Figure 2. Plan view of tunnel crossing bridge foundation.

26900
410
20 6000
to Tianshui Road
40
6980
6980
40
6000 20 410
6.000
to Chifeng Road
5.940
5.000
4.020
1
backfilling
2.500
+1.16
2 3-1 grey clay silt
13688
14-400 400 27000

quadratic precast pile
-7.69
EPB shield machine
2 3-2 grey sandy silt
-10.04
PROJECT OVERVIEW
grey silt soil
-13.24
5 1-1 grey caly soil
-17.24
2.1
5 1-2 grey silt clay
Status of Shajinggang Bridge
-21.04
-22.700
According to the plan of Shanghai subway Line 10,

the interval tunnel from Liyang Road to Quyang Road
will cross through the piles of Shajinggang Bridge
on Siping Road. The bridge is a simple beam structure with three spans; its spans are 6 m, 12 m and 6 m
respectively. This bridge includes two piers and two
abutments. Each pier uses 23 quadratic reinforced
concrete piles as its foundation, with dimension of
400 400 26000 mm; while each abutment uses 14
quadratic reinforced concrete piles as its foundation,
with dimension of 400 400 27000 mm. Since the
elevation of tunnel vault beneath the bridge is about
6 7 m, in the process of shield drive, there are
about 4 piles at each pier and 3 4 piles at each abutment will affect tunneling excavation, as shown in
Figure 2 and 3.
Around Shajinggang Bridge, there are intensive
buildings and underground pipelines, which should
be taken into account during the construction process.
The important buildings near the bridge are a high-rise
building in the southwest, a single story pump station
in the south, a 24-storey building in the southeast, a
reinforced concrete and brick building with 4 floors in
the northeast, the pit of Quyang Road Station in the
north and the Huaxi Stock Exchange Building in the
northwest. The important underground pipelines are a
group of power cable, a water supply pipe with diameter of 1500 mm and a group of telephone cables to the
east of the bridge; while two gas pipes with diameter of
300 mm and 700 mm respectively, a water supply pipe
with diameter of 300 mm and two groups of telephone
cables to the west of the bridge.
-25.24
sap green ~ olive drab grey silt clay
71
olive drab ~ grey sandy silt
72
olive ~ grey silt
-22.180
23-400 400 26000

quadratic precast pile
-32.69
Figure 3. Elevation view of tunnel crossing bridge

foundation.
2.2 The geological conditions of the project

According to engineering geological survey report, the
geological conditions at Shajinggang Bridge can be
divided into the following layers. The first layer is
miscellaneous backfilling with thickness from 1.9 to
3.7 m, containing cinder, stone, etc. The second layer
is grey and yellow clayey mingled with 1.0 m thick
black humus in it. The third layer is grey silt soil containing clay in it. It belongs to medium compression
soil, about 10 m thick. The fourth layer is grey clay
with sand body trapped in it. It is high compression
soil, including white stiff block and black humus in
it, about 5 12 m thick. The fifth layer is dark green
hard clayey. It belongs to medium compressed soil,
about 2.4 2.8 m thick. The sixth layer is olive drab
hard clay, which is medium compression soil also. The
seventh layer is olive drab silt soil.
478
2.3
Existing problems and solutions
In this project, the uniqueness and difficulties is not

only reflected on complicated site topography and
more intensive surrounding buildings, but also can be
seen from the following aspects.
1. Since Quyang Road subway station is very near to

the bridge pile foundation, coupled with the restriction of line slope in longitudinal section, the tunnel
can not cross beneath the pile foundation directly.
2. Since the outer diameter of shield machine is
approximately 7 m, in the crossing process, 4 piles
at each pier and 3 piles at each abutment should be
removed.
3. Since Siping Road is one of main roads of Shanghai
city, its traffic flow is big at day and night, it is not
allowed that the traffic volume to be affected during
the construction period.
Portal-type frame
Cutted pile sector
Pile
Casing pipe
Figure 4. Sketch map of deep excavation.

chain
casing pipe
steel bar
cutting position
truncated pile
In view of the above considerations, under the

premise of existing traffic not to be affected, the proposed construction strategies include: adopting pile
foundation underpinning technology at each pier and
abutment; removing piles affecting shield tunneling. Since Shanghai has not yet had the precedent of
pile under-pinning in subway construction, it is necessary to conduct series comprehensive technical studies
on this technology and its accessory method, thus provide a common technical guidance for future similar
projects.
3 ACCESSORY CONSTRUCTION METHODS
IN PILE UNDERPINNING
3.1
Methods of foundation reinforcement
The purpose of foundation reinforcement is to ensure

the stability of existing structures, such as pier and
abutment, and to provide stable foundation.
The main methods of foundation reinforcement can
be concluded as:
1. Grouting method: the main role of grouting is to
seal water, but it is not applicable to cohesive soil.
2. Deep mixing method: it can ensure foundation intensity and its sealing characteristics. This
method includes two categories, that is, vertical
construction method and transverse construction
method.
3. Freezing method: it can guarantee foundation
strength and sealing effect; but it is difficult to
resolve the ground deformation when defrosting.
3.2
Method of removing existing piles
There are two methods, that is, extracting piles from

ground surface and demolishing piles within layers.
(1) Extracting piles from ground surface
Deep excavation method: it is very simple, cheap
and the required operation space is small also. The
main procedure is shown in Figure 4.
Extracting piles with casing pipe: according to some
overseas related experiences, this method includes
processes of extracting piles with casing pipe and
pile
(a) Uplifting the pile
(b) Cutting the pile
(c) Removing the pile
Figure 5. Sketch map of extracting pile with casing pipe.
cutting piles, only needing 5 6 m construction space

in height, as shown in Figure 5.
(2) Demolishing piles within layers
This method can also be divided into two types:
demolishing piles manually inside the chamber and
cutting piles directly by shield machine. For the first
type, it should reinforce the foundation firstly, and then
remove piles manually in the chamber. For the sake of
safety, compressed air can be used as supplementary
measure to stabilize the excavation face. For the second
type, piles can be cut directly by the special device
installed on the cutter disc of shield machine without
human operation. But, the shield machine used here is
relatively more expensive.
4
PROPOSAL FOR SHAJINGGANG BRIDGE
Based on overseas experience and some successful

instances [4], two feasible schemes are provided here.
A scheme is reinforcing foundation and eliminating
piles inside the shield chamber, while B scheme is reinforcing foundation, adding new piles and extracting
piles with casing pipe.
4.1 A scheme
(1) Foundation reinforcement and construction of
retaining wall
Foundation reinforcement and construction of
retaining wall is to form operation space which is used
to reinforce bridge structure, as shown in Figure 6 and
7. Since the river flows perennially under the bridge,
cofferdam should be built firstly.
(2) Pit excavation behind the abutment
Since the foundation of pier and abutment can not
be reinforced on road surface, so it is necessary to
479
foundation reinforcement behind retaining wall

retaining wall
foundation reinforcement
of pier and abutment
backfilling
foundation reinforcement behind abutment
foundation reinforcement outside the tunnel
Figure 6. Elevation view of foundation reinforcement.
Figure 9. Elevation
reinforcement.
foundation reinforcementbehind abutment

retaining wall
view
of
structure
foundation
backfilling
foundation reinforcement behind abutment
Figure 7. Plan view of foundation reinforcement.
foundation reinforcement of pier
steel plate
Figure 10. Foundation reinforcement of pier.
4m
backfilling
excavation
behind
abutment
Figure 8. Elevation view of pit excavation behind abutment.
dig pit between the abutment and retaining wall. It is

noteworthy that the excavated depth below the road
surface should be no less than 4 meters so as to reinforce the foundation smoothly. After excavation, the
pit top should be covered with steel plate to ensure
that the ground traffic will not be affected, as shown
in Figure 8.
(3) Foundation reinforcement of pier and abutment
Once the foundation of pier and abutment had been
reinforced, as shown in Figure 9 to 11, it can not
foundation reinforcenment of abutment
Figure 11. Foundation reinforcement of abutment.
only transfer the load borne by the removed piles to

the new added piles and the remaining piles, but also
integrate the foundation and piles as a whole body.
In addition, the tunnel lining located here should be
redesigned as high stiffness segment to withstand the
load transferred from the removed piles. Considering
480
add new piles
Structure reinforcement of pier and abutment
foundation reinforcement outside the tunnel
Figure 14. Elevation view of foundation reinforcement.

Figure 12. Elevation view of reinforcing pier and abutment.
length below the bridge, which is indicated by the purple line in the figures. Meanwhile, the deformation of
pier and abutment should be monitored strictly.
concrete layer
(6) Maintaince of steel beam and restoration of road

surface
compressured air method
high stiffness segment

foundation reinforcement
outside thetunnel
Figure 13. Demolishing piles and propelling shield

machine.
To prevent the corrosion of steel beam, it should

be coated with a layer of concrete. After the obstacle
piles had been removed and the shield machine had
passed by, in order to guarantee regular transportation
running, it is necessary to demolish the retaining wall,
backfill the pit and restore the road surface. But, during these processes, more attention should be paid to
monitor the deformation of structure.
4.2 B scheme
the security of manually operation and stability of tunneling face, the foundation outside the tunnel should
also be reinforced firstly, and the reinforced scope is
indicated by the blue line in following figures.
(4) Structure reinforcement of pier and abutment
The purposes of reinforcing pier and abutment
structure is to transfer the load borne by the removed
piles to the reinforced foundation and remainder piles,
and improve the capacity to resist shear force and
moment perpendicular to the axial of pier and abutment. During reinforcement, steel beam can be used
to support the structure, as shown in Figure 12.
(5) Demolishing piles and propelling forward the
shield machine
After all the reinforcement work had been done, it
will be able to demolish piles, propel forward shield
machine and assemble segments subsequently. Since
the work of demolishing piles is done inside the chamber, the excavated soil in it should be discharged firstly.
Thereafter, tunneling face is in the unsupported status,
so compressed air can be used to stabilize the face. The
whole sketch map is shown in Figure 13. If necessary,
high stiffness segment may be used within the whole
(1) Foundation reinforcement and the construction of

retaining wall
Foundation reinforcement and the construction of
retaining wall are the same as that in A scheme, as
shown in Figure 6 and 7.
(2) Pit excavation behind the abutment
This process is also identical with that in A scheme,
as shown in Figure 8.
(3) Foundation reinforcement of pier and abutment
and the construction of adding new piles
Compare with that in A scheme, there are two differences in this process altogether, first is the reinforced
scope, second is adding new piles, as shown in Figure 14 to 17. After the foundation reinforcement, the
foundation at the topside around the piles should be
excavated to form working space for adding new piles.
Therefore, the required vertical clearance under pier
and abutment should be not less than 3.5 meters.
(4) Reinforcement of pier and abutment structure
Reinforcement of pier and abutment can better
improve its capability to resist exterior loads. Therefore, steel beam can be used to support the structure;
481
new piles
pit excavation behind abutment

steel cover ploate
steel beam
foundation
reinforcement
retaining wall
new piles
Figure 15. Plan view of foundation reinforcement.
foundation
reinforcement
existing piles
Figure 18. Reinforcing bridge structure and adding new

piles.
backfilling
pit excavation, pile extraction
5m
Figure 16. Foundation reinforcement of pier.

Figure 19. Further excavation of foundation.
backfilling
backfilling
foundation reinforcement ofabutment
removed piles
Figure 17. Foundation reinforcement of abutment.
Figure 20. Extracting pile with casing pipe at pier.
the main process is identical to that in A scheme, as

shown in Figure 18.
(5) Foundation excavation and pile extraction
The purpose of further excavation of foundation
below the bridge is to ensure that the working space for
extracting pile can be satisfied. But, only those places,
where the piles need to be extracted, should be excavated down. Since casing pipe method is used to extract
piles, the required vertical clearance under the bearing

platform should be no less than 5 m, as shown in Figure 19 to 21. During the construction of extracting pile,
more attention should be paid on the monitoring of pier
and abutment, especially for structure deformation.
(6) Propelling shield machine and installing segments
After the impeded piles had been removed, the
shield machine can then move forward and pass
482
backfilling
removed piles
foundation
reinforcement of abutment
Figure 21. Extracting pile with casing pipe at abutment.
backfilling
demolishing the retaining wall, restoring

the ground surfacement
CONCLUSIONS
Combined with the practical situations of in Shanghai area and the example of shield machine crossing
through the pile foundations of Shajinggang Bridge,
this paper mainly study how the underpinning technology can be used in subway construction. Based
on concrete analytical studies, two main construction
schemes are presented here. One is reinforcing foundation and eliminating piles inside the chamber; the
other is reinforcing foundation, adding new piles and
extracting piles with casing pipe. Since the pile underpinning technology has never been used in the subway
construction of Shanghai area, the two schemes provided in this paper can not only be used as reference of
design and construction, also as precious experience
for future similar projects.
REFERENCES
Figure 22. Backfilling and restoration of road surface.
through bridge foundation, followed by installation of

segments. During the crossing process, deformation
of pier and abutment structure should be monitored
primarily.
Ishimura, T., Metoki, M. & Shimizu, M. 2006. Development of removed pile method with cutting. Tunnelling and
Underground Space Technology 21: 411412.
Iwasaki, Y., Watanabe, H., Fukuda, M., Hirata, A. & Hori, Y.
1994. Construction control for underpinning piles and
their behavior during excavation. Geotechnique 44(4):
681689.
Sharma, J.S., Hefny, A.M., Zhao, J. & Chan, C.W. 2001.
Effect of large excavation on deformation of adjacent
MRT tunnels. Tunnelling and Underground Space Technology 16: 9398.
Yamaguchi, I., Yamazaki, I. & Kiritani, Y. 1998. Study of
ground-tunnel interactions of four shield tunnels drive in
close proximity, in relation to design and construction
of parallel tunnels. Tunnelling and Underground Space
Technology Vol.13, No.3: 289304.
(7) Backfilling and restoration of road surface

Once the shield machine had passed through the
pile foundation, it is necessary to demolish the retaining wall, backfill the pit and restore the road surface
to guarantee the regular traffic running, as shown in
Figure 22. During such process, it is important to
put emphasis on the deformation monitor of bridge
structure.
483
Characteristics of tunneling-induced ground settlement in groundwater

drawdown environment
C. Yoo & S.B. Kim
Sungkyunkwan University, Suwon, Korea
Y.J. Lee
RIST, Yongin, Korea
ABSTRACT: This paper presents the results of an investigation on the characteristics of tunnelling-induced
ground settlement in groundwater drawdown environment. The dynamics of the effect of groundwater drawdown on the ground settlements are first investigated using a case history concerning a conventional tunnelling
situation in which the interaction between the tunnelling and the groundwater induced excessive ground surface
settlements. A 2D stress-pore pressure coupled finite element analysis is then conducted on a tunnelling case
with groundwater drawdown, aiming at investigating ground surface settlement characteristics. The results indicated among other things that significant portion of ground settlement can occur before tunnel face reaches, and
that the error function approach does not provide a good fit to the settlement troughs for tunnelling cases with
groundwater drawdown.
INTRODUCTION
Tunnelling beneath the groundwater table causes

changes in the state of stress and the pore water pressure distribution. In such tunnelling problems, the
tunnelling work inevitably causes water inflows into
excavated area, thus causing the change in the pore
water pressure distribution. The direct environmental consequence of water inflows during tunnelling is
the drawdown of groundwater level in the surrounding aquifer (Yoo 2005). The related ground subsidence
occurring as a result of the reduction in water pressures
in the soil layers can damage nearby structures/utilities
(Figure 1). One of the major case histories illustrating damage due to ground settlement associated with
tunneling-induced groundwater drawn is perhaps the
Romeriksporten tunnel in which the highspeed railway tunnel construction caused more 1 m of ground
subsidence due to groundwater drawdown, raising significant technical and political issues pertaining to
the effect of tunnelling on surrounding environment
(NSREA 1995).
Surprisingly studies concerned with the tunnelinginduced ground movements in groundwater drawdown
are scarce as indicated by Yoo (2005). Although a
number of studies on the ground subsidence caused
by groundwater pumping from an aquifer have been
conducted (Shen et al. 2006; Qiao & Liu 2006;
GWT
Ground
surface
Groundwater
lowering
Reduction in
pore pressure
Figure 1. Illustration of ground settlement associated with

tunnelling-induced groundwater drawdown.
Xu et al. 2006). The results of their studies cannot be directly applied to the tunnel excavation
problems as they focused only on the groundwater
drawdown due to groundwater pumping. As urban
tunnelling projects tend to involve potential problems related to groundwater drawdown ground movements during tunnelling, there is an urgent need for
better understanding on the mechanisms involved
in tunnelling-induced ground movements associated
with groundwater drawdown.
This paper presents the results of investigation on
the characteristics of tunnelling-induced ground settlement in groundwater drawdown environment. The
485
Figure 2. Typical ground profile.

Table 1.
Type
fill
alluvial
weathered
soil
weathered
rock
hard rock
Geotechnical properties of soil/rock layers.
c

E
(kN/m3 ) (kPa) (deg) (MPa)
k
(cm/sec)
18
20
25
0
15
15
27
30
30
5
10
50
0.40 3.8 104

0.40 3.8 104
0.33 2.4 104
25
60
35
120
0.30 8.8 105
26
100
35
200
0.25 5.0 105
Figure 3. Support pattern (typical).
2.2 Tunnel design
Note: = unit weight, c = cohesion, = internal friction angle, E = youngs modulus, = poissons ratio,
K = coefficient of permeability
dynamics of the effect of groundwater drawdown on

the ground settlements are first investigated using a
case history concerning a conventional tunnelling situation in which the interaction between the tunnelling
and the groundwater induced excessive ground surface settlements. A parametric study using a calibrated
2D stress-pore pressure coupled finite element model
is then conducted on a number of factors influencing the tunnelling-induced ground settlements with
groundwater drawdown. Based on the results, the interaction mechanism between the tunneling, groundwater
lowering, and ground settlement is identified.
GROUND SURFACE SETTLEMENT

CHARACTERISTICS FIELD MONITORING
DATA
2.1 Tunnelling condition

A case history concerning the conventional tunnelling,
i.e., NATM, was considered. The tunnel has excavation
width and height of approximately 10.5 m and 8.7 m,
respectively, with a cover depth ranging approximately
20 30 m, and constructed in a multi-layered ground
including a fill, alluvium, and a weathered zone as
illustrated in Figure 2. The geotechnical properties of
the ground are given in Table 1.
Figure 3 shows a typical tunnel support pattern used

for a 100 m long section for the ground profile given
in Figure 2. On account of the difficult ground condition the ring cut excavation method was adopted to
promote the tunnel face stability during excavation.
The primary support system consisted of a 0.2 m thick
steel fibre reinforced shotcrete (SFRS) layer with 4 m
long system rock bolts at 1.0 and 1.2 m, respectively,
longitudinal and transverse spacing. The pipe umbrella
technique using 800 mm diameter grout injected 12 m
long steel pipes was additionally implemented to promote the face stability through improving the load
carrying capacity of the ground ahead of the face. Also
adopted was a trumphet shaped micro cement injection (MSG) pre-grouting around the tunnel periphery
to create a 5 m thick watertight shell for sections in
which the weathered soil layer extended to the tunnel crown level. The pre-grouting scheme was later
extended to cover the face after the settlement problem
had become an issue.
2.3 Measured ground surface settlements
Figure 4 shows the progressive development of settlements during the tunnel advancement at various
monitoring stations. In this figure the measured settlements are plotted against the relative distance between
the tunnel face and the monitoring stations normalized by the tunnel diameter (D). These data, measured
using the conventional leveling technique, thus indeed
represent the settlement history during the tunneling
process for the monitoring stations.
A total of five settlement curves are presented in
this figure. As can be seen in this figure the five curves
are similar both in qualitative and quantitative terms,
despite the vertical extent of the decomposed soil relative to the tunnel varies along the route, showing
the maximum converged settlements in the range of
486
of the final settlement (Sv,max ) was completed before

the face passed a monitoring station with the remaining 30 40% occurred after the full passage of the
tunnel face. Such a percentage of settlement ahead of
the face is considerably larger than the typical value of
40 50%, suggesting a larger portion of the converged
settlement occurred prior to the arrival of tunnel face
in this tunneling condition than a tunneling condition
without the groundwater drawdown. Moreover, the settlements tend to converge to a constant value after the
tunnel face advanced to a distance of 6 7D beyond
the monitoring stations, suggesting slower settlement
convergence than a normal condition.
These results in fact are somewhat different from
typical trends that can be observed in tunneling conditions without significant groundwater lowering, and
led to a conclusion that factors other than the unloading
effect due to the tunnel excavation may have played a
role. Such a tendency is directly linked to the groundwater drawdown as will be shown in a subsequent
chapter.
Figure 4. Progressive development surface settlements at

various monitoring stations.
PARAMETRIC STUDY
3.1 Stress-pore pressure coupled analysis
Figure 5. Normalized settlement history curves.
1.6% 1.8%D. Of salient feature that can be observed

in this figure is the tendency of settlement increase
during which the tunnel advancement was halted, suggesting time dependent (tunneling activity independent) settlement development (to be discussed later).
Another of interest trend is the resemblance of the
settlement curves with a typical log t curve from a consolidation test, being characterized by three zones as
an initial compression, primary, and a secondary zone.
Such a trend strongly suggests a possible cause being
the volume change effect due to tunneling-induced
groundwater drawdown.
The data in Figure 4 are further analyzed by normalizing the settlement values with their respective
maximum values (Sv,max ) in Figure 5. As seen in this
figure, the normalized curves tend to collapse into one
curve. A further inspection of the normalized settlement history curves shows that the settlements started
to develop when the tunnel face was approximately 6D
away from the monitoring stations. The settlements
tend to accelerate when the tunnel face reached 3D
away from the monitoring stations, and decelerate after
the tunnel advanced 5 6D beyond the monitoring
stations. Also shown are that approximately 60 70%
A commercial finite element package ABAQUS

(Abaqus, Inc. 2002) was used for the parametric study.
A 2D stress-pore pressure coupled effective formulation was adopted in order to realistically capture the
interaction mechanism between the tunnelling and the
groundwater.
In ABAQUS a porous medium is approximately
modelled by attaching the finite element mesh to
the solid phase. Equilibrium is expressed by writing
the principle of virtual work for the volume under
consideration in its current configuration at time t:
where dv is a virtual velocity field, d is the virtual

rate of deformation, is the true (Cauchy) stress, t are
surface tractions per unit area, and f body forces per
unit volume. f includes the weight of the wetting liquid
fw defined as Eq. (2)
in which s is the degree of saturation, n is the porosity,

and nt is the volume of trapped wetting liquid per unit
of current volume. Eq. (1) can then be rewritten as
487
Table 2.
Geotechnical properties.
soil
Weathered
rock
(kN/m3 )
E
(kPa)
c
(kPa)
( )
k
(cm/sec)
25
25
50,000
100,000
30
50
30
30
5.8 103
1.3 104
Figure 6. Finite-element model used in the analysis.
where f are all body forces except the weight of the

wetting liquid.
The continuity equation is satisfied approximately
in the finite element model by using excess wetting
liquid pressure as the nodal variable (degree of freedom 8), interpolated over the elements. The backward
Euler approximation is used to integrate the equation over time and the Newton iterations are used to
solve the nonlinear, coupled, equilibrium and continuity equations. Fundamentals of the stress-pore pressure
coupled formulation adopted in ABAQUS can be
found in the ABAQUS users manual (Abaqus, Inc.
2005).
3.2
Condition analyzed
A tunnelling condition frequently encountered in

urban situations was considered in the analysis. The
tunnel considered is a 10 m diameter horseshoe shaped
tunnel with a cover depth of 3.0D, excavated by the
bench cut method. The primary support system consists of a 20 cm thick shotcrete lining with system
rock bolts installed at 1.5 m center-to-center spacing.
A 1.5D thick soil layer was assumed to exist above a
weathered rock layer through which the tunnel is excavated. Tables 2 summarizes geotechnical properties of
the ground.
3.3
The ground and the shotcrete lining were discretized

using 8-node displacement and pore pressure elements
with reduced integration (CPE8RP). The rock bolts
were modeled using the 2-node truss elements. With
regard to the material modelling, the soil and rock
layers were assumed to be an elasto-plastic material
conforming to the Mohr-Coulomb failure criterion
together with the nonassociated flow rule proposed by
Davis (1968), while the shotcrete lining and the rock
bolts were assumed to behave in a linear elastic manner. The geotechnical properties for the ground given
in Table 1 were used for analysis. The youngs moduli
of the shotcrete and rock bolts were chosen as 15 GPa
and 21 GPa, respectively.
The actual tunnelling process consisting of a series
of excavation and support installation stages was
closely simulated in the analysis by adding and removing corresponding elements at designated steps. After
establishing the initial stress and pore pressure conditions with appropriate boundary conditions, the stepby-step tunnelling process pertinent to the bench cut
excavation method, was then simulated. The 3D effects
of advancing a tunnel heading was taken into consideration using the stress relaxation method in which the
boundary stresses arising from the removal of excavated elements are progressively applied to simulate
the progressive release of the excavation forces as the
tunnel heading advances.
Finite element model
Figure 6 shows the finite element model adopted in

this study. The finite-element mesh extends to a depth
of two times the tunnel diameter (D) below the tunnel spring line and laterally to a distance of 15D from
the tunnel center depending on the cover depth H . The
lateral location was selected based on a series of preliminary analysis as it has a significant influence on
the results of a stress-pore pressure coupled analysis.
At the lateral boundary displacements perpendicular to
the boundaries were restrained whereas pin supports
were applied to the bottom boundary.
With regard to the hydraulic boundary conditions
and with reference to Figure 6, a no-flow condition
was assigned to the vertical boundaries perpendicular
to the tunnel drive. At the lateral vertical boundary the
groundwater table was assumed to be at the ground
surface and constant throughout the analysis.
3.4 Ground settlement characteristics

Figure 7 presents the relationship between the maximum surface settlement (Sv,max ), directly above the
tunnel crown, obtained during various stages of tunneling. As seen the maximum surface settlement
Sv,max tends to linearly increase with the increase in
the groundwater drawdown level HD . The settlement
occurred after the completion of tunnel is in fact
twice that during the tunnel excavation. It should be
noted that the plot given in this figure represent those
caused by the groundwater inflow into the tunnel after
the completion of tunnel excavation, until a steady
state condition is achieved. This suggests a direct link
between the ground settlement and the groundwater
drawdown, thus demonstrating the importance of creating a watertight shell for tunnelling cases where the
controlling ground surface settlement is of concern.
488
Figure 7. Variation of Sv,max with HD after completion of

tunnel excavation.
Figure 9. Normalized surface horizontal displacement profiles.
tend to collapse into one curve despite some discrepancies in the region father away, i.e., 4D, from the
tunnel center.
Normalized horizontal displacement (Sh ) profiles
are shown in Figure 9 for different levels of groundwater drawdown. As seen, the maximum horizontal
displacements (Sh,max ) tend to develop at locations
3D away from the tunnel center with decreases in
magnitudes thereafter. Again the Sh profiles tend to
collapse into one curve although some discrepancies
are observed in the region away from the tunnel center.
The results obtained in this study suggest that the settlement trough as well as the horizontal displacement
profile may be constructed using normalized curves
when relationships between the surface movements
and other factors can be established.
Figure 8. Normalized surface settlement troughs.
In urban tunnelling situations, characteristics of

a ground surface settlement trough, such as slope
and width of inflection point, are important as they,
together with lateral displacements, determine potential for damage of adjacent structures. Figure 8 shows
normalized ground surface settlement troughs for different groundwater drawdown levels together with
error functions constructed using different values of
inflection point. Of interest trends are two fold. First,
the extent of ground settlement trough is significantly
greater than typical tunnelling conditions without
groundwater drawdown. In fact, for the particular tunnelling condition considered, the ground settlement
zone extends more than 10 times the tunnel diameter
from the tunnel centerline. Second, the error function
approach (Attewell et al. 1986; Peck 1969) known to
well describe the surface settlement trough for tunnelling cases without ground water drawdown does
not provide a good fit to the settlement troughs for
cases with groundwater drawdown. Another important
observation is that the computed settlement troughs
CONCLUSIONS
This paper presents the results of an investigation on

the characteristics of tunnelling-induced ground settlement in groundwater drawdown environment using
the measured surface settlement for a site where
the tunnelling-induced groundwater drawdown caused
significant surface settlement. A stress-pore pressure
coupled finite element model was additionally conducted aiming at identifying the ground movement
characteristics when tunnelling induces a significant
level of groundwater drawdown. Based on the results
the following conclusions can be drawn.
1. For tunnelling cases in which tunnel excavation
causes significant groundwater drawdown, the percentage of settlement that develop prior to the
tunnel face arrival to the final settlement is significantly larger than for cases without groundwater
drawdown.
489
2. Continued groundwater drawdown after the completion of tunnel excavation may can cause settlement larger than that occur during excavation.
3. The error function does not provide a good fit to
the settlement troughs for cases with groundwater
drawdown.
4. Normalization can hold for the surface settlement
and horizontal displacement profiles for tunnelling
cases with groundwater drawdown.
ACKNOWLEDGEMENT
This research is supported by Korea Ministry of Construction and Transportation under Grant No. C4-01.
The financial support is gratefully acknowledged.
REFERENCES
Abaqus users manual, Version 6.5. 2005. Hibbitt, Karlsson,
and Sorensen, Inc., Pawtucket, Providence, R.I.
Attewell, P.B., Yeates, J. & Selby, A.R. 1986. Ground deformation and strain equations. Soil movements induced by
tunneling and their effects on pipeline and structures,

Blackie, Glasgow: 5366.
Norwegian Soil and Rock EngineeringAssociation (NSREA).
1995. Norwegian urban tunnelling. Publication No. 10.,
Norway.
ground Proc., 7th Int. Conf. on Soil Mech. And Found.
Engrg. : 225290.
Qiao, S. & Liu, B. 2006. Prediction of ground displacement
and deformation induced by dewatering of groundwater. Underground Construction and Ground Movement:
7379.
Shen, S.L., Tang C.P., Bai, Y. & Xu, Y.S. 2006. Analysis of
settlement due to withdrawal of groundwater around an
unexcavated foundation pit. Underground Construction
and Ground Movement: 377384.
Xu, Y.S., Shen, S.L. & Bai, Y. 2006. State-of-art of land
subsidence prediction due to Groundwater Withdrawal
in China. Underground Construction and Ground Movement: 5865.
Yoo, C. 2005. Interaction between Tunnelling and
Groundwater-Numerical Investigation Using Three
Dimensional Stress-Pore Pressure Coupled Analysis.
Journal of Geotechnical and Geoenvironmental
490
Effect of long-term settlement on longitudinal mechanical performance of

tunnel in soft soil
H.L. Zhao, X. Liu, Y. Yuan & Y. Chi
ABSTRACT: This paper introduces the method for analyzing settlement data of tunnel in soft soil. The method
is applied to the settlement monitoring analysis of some tunnels across the river. In the case study, the longterm monitoring data for shield tunnel settlement is discussed concretely. Firstly, unitary longitude settlement
curves are gained, and the corresponding longitudinal curvature is calculated to investigate the rules of absolute
settlement and uneven settlement varying from time. Secondly, based on the uneven settlement magnitude, the
mechanical properties of shield tunnel are discussed by means of the equivalent axis stiffness continuous model
of shield tunnel. In this way, the relationships between longitudinal curvature and crack of shield tunnel segment
joint, the stress of tunnel concrete segment, and the stress of bolts for segment joint are found generally. So
the crack and leakage of shield tunnel can be inferred consequently. Finally, the results gotten from analytical
solution are compared with the field inspection information, and the favorable agreement is shown. Analysis of
effect of long-term settlement on longitudinal mechanical performance of tunnel in soft soil offers guidance to
evaluate the safety and serviceability, structural protection and waterproof design of shield tunnel in soft soil.
INTRODUCTION
The dramatic longitudinal settlements occur in soft soil

after many yearsoperation because of a number of factors, including vibration induced by trains, leakage,
geotechnical conditions, and depth of embedment,
construction methods, surface loads and mutative
environment actions.
Due to the great span of time and huge inspecting
cost, the investigation for tunnel long-time settlements
has been limited. By far, there are few special reports
for long-term settlement research of the tunnel. All
of these factors restrict the development of safety
and serviceability evaluation for tunnels in long-term
operating process.
The settlement reflects the longitude deformation
of tunnels. Generally, the tunnel in soft soil is weak
relatively to resist deformation caused by extraneous
factors, and the crack of circumferential segment seam
is prone to appear, which leads to leakage and segments
damage under longitudinal tension or compression.
Therefore, its crucial and significant to investigate
and analyze some rules for shield tunnel such as the
relationship of the segment inner force caused by
uneven deformation, crack of segment joint, and stress
of bolts.
In this paper, the tunnel longitude deformation curvature is obtained from monitoring data of long-term
settlement. Then, the relationship between longitude
deformation curvature with the segment inner force
and crack of segment joint are concluded. This paper
aims at evaluating the performance of tunnel after
many years service according to the corresponding
longitude deformation curvature. The analysis results
offer important foundational materials for structural
safety and durability evaluation of shield tunnel.
2
LONG-TERM SETTLEMENT DATA

ANALYSIS FOR SHIELD TUNNEL
This paper focuses on a highway tunnel across the river.

This tunnel located in soft soil has serviced for more
than 30 years. Tunnel settlement monitoring has being
executed in the service stage. In the settlement monitoring, there are two elevation benchmarks set in both
ends of tunnel. Settlement monitoring points between
the two base points are arranged in both sides of way
respectively along tunnel length, so it is a closed circuit
style for inspection.
The investigation of this paper is based on the
settlement data of this tunnel for recent ten years.
491
Figure 1. Longitudinal settlement curve of tunnel.
Figure 2. Longitudinal curvature of tunnel from Shaft No. 4 to No. 2
2.1
Longitudinal settlement curve
There is certain rigidity for the structure between

two adjacent monitoring points, and local structural
deformation will affect adjacent structural deformation and internal force. In order to reflect integral
tunnel longitude settlement comprehensively, the settlement analysis should adopt unitary analysis method.
A unitary longitude settlement curve can be drawn
based on monitoring results, shown as Figure 1. It
should be noted that monitoring points located in both
sides of this tunnel. So the settlements of these monitoring points are termed as settlements of left line and
right line respectively in this paper.
The varying rules of longitudinal settlement
obtained from longitudinal settlement curve can be
concluded as follows.
1. Most settlement magnitude of observation points is
between 30.0 mm and 30.0 mm. The annual settlement curve is parallel approximately. It means
that the increase of settlements of observation
points is uniform and tunnel longitude deformation shows favorable integrity. The deformation
difference is little between adjacent points, and they

almost rise and fall together.
2. The settlement magnitudes are variable from time.
Most settlement points oscillate up and down. For
instance, the accumulative longitudinal settlement
in 2004 is larger than that in 2002, but the settlement
in 2006 is less than that in 2004. Some of these
settlements increase continuously in the oscillating
and some incline towards steady.
2.2 The curvature of longitudinal settlement curve
The curvature change is the important characteristic
of longitudinal settlement for shield tunnel, because it
demonstrates the uneven settlement of tunnel.
The settlement data from tunnel longitude observation points is made to spline interpolation fitting, and
then, based on the interpolation result, the curvature
is elicited by numerical differential. Therefore, the
curvature varying from longitudinal distance can be
drawn, shown as Figure 2. The curvature varies from
tunnel longitudinal distance and the increased range of
curvature is large. This is in line with the rule of tunnel
492
Figure 3. Longitudinal curvature of right and left line from Shaft No. 2 to No. 1
longitude settlement development. But it should be

noted that the curvature change of left curve is different
from right curve, so it will lead to lateral deformation
and lateral additional inner force.
1. The curvature of settlement curve for the points
from No. 4 to No. 2 shaft is small. But the curvature
where closed to shaft is larger than that between
shafts.
2. The curvature is very large from No. 2 to No. 1
shaft, so another figure is drawn to describe it. The
average value of curvature is 0.01 mm-1, even up to
0.021 mm-1, shown as Figure 3. So the structural
safety and durability of this part tunnel should be
calculated and evaluated in detail.
3
SHIELD TUNNELING LONGITUDINAL

STRUCTURE MODEL
At present, there are two main kind of theoretical

research for tunnel longitude structure. One is numerical solving model based on the finite-element method.
And the other is theoretical solving model.
Japanese scholars proposed two kinds of tunnel
longitude structure theory based on the different simplification for tunnel segments joints and bolts. One
is beam-spring model (Zhu, 1998); the other is equivalent continual rigidity model (Shiba & Kawashima,
1988). The study herein adopts the equivalent continual rigidity model proposed by Shiba and Kawashima
(1988) to analyze tunnel longitude performance. Considering the tunnel analyzed in this case study has
structural lateral deformation caused by long-term
operation, so a relative correction factor is introduced to reduce the affection on tunnel longitude
deformation and inner force.
3.1
Figure 4. Relationship for P
Equivalent rigidity continuous model
Equivalent rigidity continuous model is supposed to

be continuous lining rings and homogeneous in the
cross section of tunnel. The rings compose a tubular
structure connected by longitudinal steel bolts termed
as girth joint of tunnel segment. The girth joint bolt
is simulated to spring with different characteristic
in compression and tension. The simulate spring is
Figure 5. Stress distribution for segments in tension and

compression.
bilinear material in tension, while is rigid absolutely

in compression. The relationship of P is shown as
Figure 4.
Herein, kj1 is elastic stiffness for individual bolt,
and kj2 is plastic stiffness for individual bolt, Py is the
elastic ultimate pressure for bolt.
The assumptions in equivalent continual rigidity
model are as follows.
1. Circumferential inhomogeneity of lining ring is not
taken into consideration. Additionally, the shear
stress and deformation in longitudinal seam are
ignored.
2. Compressive pressure is resisted by concrete segment singly, and tensile pressure is resisted by
concrete segment in share with steel bolts.
3. Plane sections remain plane and small deformation
assumptions are applicable in this model.
The major results of equivalent rigidity continuous
model referenced with the analysis of this paper are as
follows.
493
Figure 6. Stress distribution for segment in elastic state.

Figure 7. Stress distribution for segment in plastic state.
3.2
Then the equivalent elastic bending stiffness for tunnels can be expressed as follow:
Equivalent compressive and tensile rigidity
A segment with length ls is taken as a calculate

unit. Its stress distribution in tension or compression
can be seen in Figure 5. The equivalent axial tensile
and compressive rigidity is gained from deformation
compatibility and force balance conditions, shown in
Equation 1.
Elastic ultimate flexural moment for priestesses segments, defined as:
Therefore, the tunnel radius of curvature corresponding to elastic ultimate flexural moment is:
(EA)Ceq , (EA)Teq1 , (EA)Teq2 is equivalent longitude compression rigidity, the elastic and plastic equivalent
tensile rigidity respectively. Kj1 , Kj2 is elastic and plastic spring coefficient. Kj1 = nkj1 ; Kj2 = nkj2 . N , N0 and
Ny is longitude axial force for unit, pre-pressure for
joints and the elastic ultimate tension for unit respectively. Ec , Ac is Youngs modulus of concrete and area
of segment.
3.3
Equivalent bending rigidity
When the segment is in absolute elasticity state and

the tensile stress of all the bolts is less than Py, the
unit has the stress and strain shown in the Figure 6, in
which, x, indicate the position of neutral axis, where,
x = r sin .
In terms of the deformation compatibility condition
and the force balance equation, where, meets the
Equation 2:
When y = D/2, the maximum compression stress for

concrete segment can be expressed as:
When y = D/2, the maximum tensile stress for

concrete segments can be expressed as:
When y = r, the maximum tension and deformation

for joint bolts respectively expressed as:
When the flexural moment of tunnel segment is

larger than elastic ultimate flexural moment My ,
the outermost joint bolts would demonstrate plastic
494
Table 1.
Main structural parameters for shield tunnel.
Inner
diameter
(m)
Outer
diameter
(m)
Youngs
Ring modulus of
width concrete
(m)
(kPa)
8.8
10.0
0.9
3.35 107
Compression
strength of
concrete
(kPa)
Bolt
diameter
Youngs
/length/
modulus
quantity
of bolt
(mm/mm) (kPa)
2.96 104
30/580/64
characteristic. The stress and strain of segment in

which some bolts reach the plastic state is shown
in Figure 7. and demonstrate the critical position where the deformation of bolts is elastic ultimate
deformation y , and = r sin .
In terms of the deformation compatibility condition
and the force balance equation, an equation concerned
with and can be obtained as follows.
4.1
Where, y is elastic ultimate deformation for joint

bolts,
(Zheng, 2005)
Where 0 is pre-pressure of bolts, y is yield stress,
u is ultimate stress and Ej is Youngs modulus. R1 ,
R2 are coefficients in equation, R1 = 1 + 1Ec Ac , and
ls Kj1
R2 = 1 + 1Ec Ac .
ls Kj2
The maximum compression stress of concrete segments
The tunnel longitude deformation curvature can be

defined as follows:
Ultimate
strength
of bolt
(kPa)
Ratio of
elastic
stiffness
and plastic
stiffness ()
5.4 105
8.35 105
0.01
The relationships between tunnel longitude curvature and stress of concrete segment, joints crack, bolts
inner force and tunnel ultimate flexural moment are
established according to the conclusions drawn from
analytic solution above. It can be used to evaluate structural performance of tunnels if the longitude uneven
settlement has been known.
4
When y = r, the maximum deformation of joint bolts

max expressed as:
2.06 108
Yield
strength
of bolt
(kPa)
LONGITUDINAL STRUCTURAL
PERFORMANCE OF TUNNEL
Critical state for evaluating performance
There are several critical conditions in the development of longitudinal stress and deformation of tunnels.
In order to evaluate the safety status of tunnels, some
primary critical states will be studied in the following
contents.
1. Crack of joints. The joints crack is limited for
meeting the demands of waterproof and structure.
2. Stress of bolts. There are two key critical states,
namely, the stress of outermost bolts reaches to
yield stress and ultimate stress.
3. Compression strength of concrete. Concrete stress
in compressive area reaches flexure compression
strength of concrete segments.
4.2 The evaluation for tunnel longitudinal
structural performance
The elastic and plastic spring stiffness coefficient
may be calculated by parameters listed in Table 1.
kj1 = Ej Aj /lj = 250929.0 kN/m, kj2 = 2509.3 kN/m.
When kj1 is introduced in Equation 2, will
be 0.9895. The elastic equivalent flexural stiffness
(EI )1eq = 2.0951 109 kNm2 . Then the value of R1
and R2 are gained based on the upper parameters, as R1 = 2.38 102 , R2 = 2.44 104 . When
y = 1.82 mm, and are calculated as = 1.36368,
= 1.20615. So the corresponding curvature can be
calculated. The inner force and deformation corresponding to the curvature of longitudinal settlement
curve for several important critical states are acquired
by the analytical method stated above, listed in Table 2
in detail.
495
Table 2.
Inner force and deformation of tunnel in critical states.

Crack of
segment
joint
(mm)
Curvature
(m1 )
Radius of
curvature
(m)
Bolt stress
(kPa)
Waterproof requirement
(crack of segment joint
1 mm)
1.10 104
27300
2.97 105
1.00
2.47 104
Concrete stress reaching

Compression strength
1.59 104
18818
4.29 105
1.21
2.96 104
Outermost bolt stress

reaching yield strength
2.00 104
15000
5.40 105
1.82
3.72 104
Outermost bolt stress

reaching ultimate strength
9.94 103
302
8.35 105
83.06
8.35 105
Critical state
When the radius of curvature is greater than

27300 m, the stress of segments, tension of bolts and
meet the demands, so the tunnel is safe relatively, but
the leakage may occur. When the radius is less than
18818 m, the compression stress of concrete in compression area will exceed the compression strength of
segments. When the radius is less than 15000 m, the
outermost bolts in compression state may be into plastic state. When the radius is less than 302m, the end
outboard bolts have reached at ultimate stress state,
and may be tensile failure.
4.3
Evaluation on structural performance
As for the part of tunnel between Shaft No. 3 and

No. 4, settlements absolute value of segment between
is small relatively and floats in the range of 20.0 mm.
The radius of curve is large. So the structural inner
force and deformation are in a low level , while the
longitudinal deformation curvature is large in the area
near Shaft No.3 and the crack of segment joints can
not meet the waterproof requirements.
As for the part of this tunnel between Shaft No. 3
and Shaft No. 2, the radius of curvature is less, about
20000 m. The field inspection results show that there is
severe leakage of inner lining of the tunnel, especially
near Shaft No. 3.
As for the part between Shaft No. 2 and Shaft No. 1,
there is great tunnel longitude settlement, and the
uneven settlement develops dramatically, especially
nearby Shaft No. 2, which becomes the weak position
in the whole tunnel.
CONCLUSIONS
From the tunnel settlement curve and curvature data,

the settlement of shield segment trends to be steady
Concrete
stress
(kPa)
after many years operation. But the uneven settlement

of tunnel develops gradually.
The magnitudes of longitude deformation curvature
corresponding to several significant crucial states are
obtained. These magnitudes can be used to evaluate
the structural performance according to the tunnels
longitude settlement.
When the radius of curvature is less than 27300 m,
the waterproof requirement for crack of segment joint
does not meet the demands, and the leakage may occur.
When the radius is less than 18818 m, the compression
stress of concrete will exceed the compression strength
of segments. When the radius is less than 15000 m,
the outermost bolts in compression state may yield.
When the radius of curvature is less than 302 m, the
outboard bolts would reach ultimate stress state, and
tensile failure may occur.
ACKNOWLEDGMENTS
This work was supported by Shanghai Leading Academic Discipline Project(Project Number: B308).
REFERENCES
Liu, J. H. & Hou, X.Y. 1991. Shield tunnel. China railway
press:369395
Shiba, Y. & Kawashima, T. 1988. The evaluation of the
ducts longitudinal rigidity in the seismic analysis of shield
tunnel. In: Proceedings of the Civil Academy :319327.
Zheng, Y.L. 2005. Study on longitudinal crack of shield tunnel segment joint due to asymmetric settlement in soft
soil. Chinese journal of rock mechanics and engineering,
24(24):45524558.
Zhu, H.H. & Tao, L.B. 1998. Study on two beam-spring models for the numerical analysis of segments in shield tunnel.
Rock and Soil mechanics, 19(2):2632.
496
Theme 4: Safety issues, risk analysis hazard

management and control
Research on stochastic seismic analysis of underground pipeline based on

physical earthquake model
X.Q. Ai
J. Li
Department of Building Engineering, Tongji University, Shanghai, P.R. China
ABSTRACT: In this paper, a new approach for the stochastic analysis of finite element modeled underground
pipeline system under earthquake excitations is proposed, where a new developed physical random earthquake
model and the probability density evolution method are adopted. Based on the physical process of seismic spread,
the random earthquake model adopted not only indicates the physical relationship between the random earthquake
ground motion and several key random parameters, but also presents the probability configuration of random
seismic motions. Associating with the developed probability density evolution method for stochastic structures,
the instantaneous probability density and the evolution process for the seismic response of underground pipeline
can be analyzed numerically, and then the stochastic response can be obtained easily. With the finite element
method, the stochastic response of underground pipeline under seismic excitation can be studied effectively.
Using the proposed method, numerical example under different random conditions is investigated, showing that
the proposed method is of high application in considering random seismic excitation with random soil parameter
at the same time.
As a basic and important problem in seismic reliability research for the underground lifelines system, the
stochastic seismic response of underground pipelines
has for a long time been extensively studied, coming up with a variety of theoretical and numerical
methods applicable to engineering practice (Machida
& Yoshimura 2002; Nedjara et al. 2007). However,
on account of the calculation limitation in previous
studies, the stochastic response analysis considering
the randomness simultaneously caused by the seismic input and the structural parameters of soil-pipeline
system is still an unsolved problem.
For the non-linear stochastic structures, it is difficult to capture the accurate probabilistic information
of dynamic performance. The dynamic response of
non-linear stochastic structures was analyzed usually
by the random simulation, the random perturbation
method or by the equivalent linearization technique.
In 1986, Liu et al.(1986) started to investigate the
problem with extension of the random perturbation
technique. In 1980s, the Monte Carlo simulation is
also employed (Deodatis & Shinozuka 1988), and
some researchers studied the techniques to reduce the
considerable computational work (Spanos & Zeldin,
1998). As the alternative approaches, the equivalent

linearization technique (Klosner et al. 1992) and the
extension methods used for linear stochastic structures
such as the orthogonal polynomial expansion (Iwan
& Huang 1996), are also investigated. However, for
the dynamic response analysis of non-linear stochastic structures, the random perturbation method had
great difficulty because of the secular terms problem
and the requirement of small coefficient of variation
of the random parameters. There are some arguments
for the equivalent linearization method because of the
misleading results in some occasions. The application of orthogonal polynomials expansion seems also
unfeasible for multiple-degree-of-freedom system and
non-polynomial-form non-linearity.
Evidently, more investigations are necessary for
the dynamic response analysis of the underground
pipelines. In resent years, a newly developed probability density evolution method for stochastic structures
has been proposed. The instantaneous probability density function and the evolution against the time can be
obtained precisely (Li & Chen 2005). For the dynamic
response of stochastic structures, the solution can be
derived through solving the probability density evolution equation with an initial value condition. On
the other hand, different from the models from the
499
power spectral density function, a new physics-based

stochastic earthquake model to reflect the fundamental correlation between the critical factors and the
random earthquake motions has been proposed by
the authors (Li & Ai 2006). Several key random
parameters are based on to realize the purpose, and
a relational expression with physical background can
be constructed considering the propagation process of
earthquake motion in an engineering site. Associated
with the probability density evolution method, the proposed random earthquake model can be of comparative
advance to provide the basis for the stochastic seismic
analysis.
In this paper, based on the probability density
evolution method and the physics-based stochastic
earthquake model, a new approach for the dynamic
response analysis of the finite element modeled underground pipeline structures under random earthquake
excitations and uncertain soil parameters is proposed
and numerical example is investigated.
2
2.1
If X = t X (, t) = G(, t), based on the above probability density evolution equation, the joint probability
density function of X and will satisfy the following
equation,
PROBABILITY DENSITY EVOLUTION

METHOD FOR STOCHASTIC STRUCTURES
Principle of preservation of probability
The principle of probability conservation can be

described as, during a conservative probability transformation process, within the state space, the increment of the probability within a unit volume equals to
the inflow probability that gets across this unit.
If the ordinary differential of a state vector Y can be
expressed as,
whereY = (y1 , y2 , . . . , yn )T , G = (g1 , g2 , . . . , gn )T .

If pY (y, t) is supposed to be the probability density
of Y (t), based on the principle of probability conservation, the probability density evolution equation can
be expressed as followed,
2.2
are stochastic mass and damping matrices respectively,

which include random parameters with the rank of
n n, n is the dynamic freedom degree; U , U , U are
the displacement, the velocity, and the acceleration
vectors, respectively; f (, U ) is the nonlinear restoring force vector; and U g is the acceleration vector of
input earthquake.
Let a response vector X = (U T , U T )T , the dynamic
equation is then changed into a format of state equation
including random parameters,
Since the velocity component X i is only the function

of variable , the above equation can be re-written as
Having integral at both sides with x1 , . . . , xl1 , xl+1 ,

. . . , x2n , the probability density evolution equation will
be decoupled,

where pXl (xl , x , t) = pY (x, t)dy1 . . . dyl1 dyl+1 . . .
dy2n , is the joint probability density function of
(Xl , T )T .
When the initial displacement and the velocity are
independent of the physical parameters of the structure, the corresponding initial condition is described
as,
Probability density evolution equation of the

seismic response of stochastic structures
The dynamic equation of the nonlinear structure can

be written as,
where is the random parameter vector which represents the physical character of a stochastic structure,
and its joint probability density function is p (x); M ,C
where Xl,0 is the determinate initial value of Xl , for a

initial static structure, there is Xl,0 = 0; () is the Dirac
function; and p (x ) is the joint probability density
function of the random vector .
Solving the above partial differential equation with
initial-boundary-values, which include the decoupled
probability density evolution equation and the initial condition, the joint probability density function
500
pXl (xl , x , t) can be calculated. After the integral with

x , the probability density function of Xl (t) can also be
solved,
The probability density evolution method of the

stochastic structures is an effective approach to obtain
the instantaneous probability density function and its
evolution for the stochastic response.
2.3 Computational algorithm of probability density
evolution equation
Step 1: Disperse the value region that corresponds to
, and disperse the initial condition simultaneously;
Step 2: After dispersing, for each determinate value
of variable , a determinate dynamic analysis is carried
out and the corresponding differential item with time
of the target response X l (x , t) can be deduced;
Step 3: Using the method of finite difference to
solve the probability density evolution equation, then
the joint probability density function pXl (xl , x , t) and
the probability density function of the target response
can be calculated.
3
RANDOM MODEL OF EARTHQUAKE

GROUND MOTION FOR ENGINEERING
SITE
Using a random Fourier function with adherent probability as the modelling form, the proposed model is to
reflect the intrinsic relationship between the random
seismic motion and the critical parameters.
3.1 The physical relation and modeling
If searching from the mechanism of the seismic
wave propagating through engineering site, earthquake ground motion can be regarded as a physical
process, which includes being input from the bedrock
base and filtered by the site. Accordingly, it is believed
that because of the uncontrollability of the random
factors including the afferent wave energy and the
soil medium characteristics that the actual earthquake
ground motions are observed with significant randomness as a result. The mentioned factors are supposed
to be the energy factor of afferent wave, the periodic
factor and the dissipative factor of seismic site. Correspondingly, these factors can be practically indicated
by the following stochastic variables with operational
physical meaning: the basal spectrum parameter, the
free angular frequency and the damping ratio of engineering site. Consequently, a physical relation between
random earthquake motion and critical factors can be
constructed and modeled.
Without loss of generality, the soil layer can be

simulated as an equivalent linear single-degree-offreedom system which is input by a seismic motion
from the base, and then the absolute response of
this system can be supposed to present the seismic ground motion process. Therefore, if this linear
single-degree-of-freedom system is input with onedimension seismic motion, in frequency domain, the
absolute acceleration response is expressed as
where is independent parameter, read as angularfrequency; Y () is the absolute output acceleration in

frequency domain; U () is the absolute basal input
acceleration in frequency domain; 0 is the natural
angular-frequency of the site; 0 is the damping ratio
of the site.
According to the previous statement, the basal input
motion and the medium characteristics of soil layer,
which lead to the randomness of earthquake acceleration motion, are both of uncertainty. Therefore, the
parameters of soil mediumthe free frequency 0 and
the damping ratio 0 are apparently stochastic variables, noted as X and X respectively. Assume random
variable vector X H = (X , X )T to relate to the soil
medium.
If the basal input Fourier spectrum U () is the function of stochastic variables X1 , . . . , Xn , U () is defined
with
Assume random variable vector presenting the basal

input Fourier spectrum to be XG = (X1 , . . . , Xn )T , and
assign random variable vector X = (XHT , XGT )T with
the joint probability pX (x). Then the random earthquake ground process described with acceleration can
be expressed by a random Fourier spectrum function
F(X, ), which is expressed as
3.2 The basal input spectrum

If presented by power spectral density function, the
bedrock seismic responses are generally assumed to
be a band-limited white noise spectrum. However, a
large amount of observed acceleration records have
shown that the bedrock spectrum presents filtering
and limited band character. Moreover, relevant analysis indicates that the frequency range greater than
501
Figure 1. Basal Input Spectrum.
Figure 3. Contrast of the proposed random earthquake

model and real records for Soil Type II.

model and real records for Soil Type I.
94 Rad/s (15 Hz) can be neglected because of corresponding minute amplitude.

According to the preceding thoughts, a basal input
spectrum is constructed (Li & Ai 2006), which is
shown in Figure 1, where, 1 , 2 are the control angular-frequency; e is the truncated angularfrequency; Xu is the random variable defining the
amplitude of the basal input spectrum.
stochastic modeling theory, actual strong seismic

records are regarded as a target sample aggregation and
numerical methods are adopted to identify different
distribution parameters of the variables corresponding to the proposed model (Li & Ai 2006). According
to different site class, a collection of acceleration
records, which chiefly come from western American
strong seismic records, are collected and reorganized
to establish different equivalent random earthquake
model.
For different site class, the contrasts between actual
seismic records and the proposed model are shown
in Figures 25. The contrasts imply that the proposed model is shown of definite physical concept and
propriety to reflect the variation of random seismic
motion.
STOCHASTIC SEISMIC ANALYSIS OF

UNDERGROUND PIPELINE
4.1 Computational example

3.3 Random earthquake model basing on physical
process
In order to determine the physical variables in the
proposed random earthquake model, according to
The site is a 50 10 m uniform saturated sandy soil

space and the ground water is set at the ground surface.
And the pipe made of cast iron is buried at the depth
of 1m with the diameter is 0.4 m.
502
Soil
Pipeline
Contact surface
10M
50M
Boundary condition
Figure 6. The Structural Mode.

model and real records for Soil Type III.
Figure 6 shows the finite element mode of the

computational example. The structure is subjected to
earthquake excitation in the shape of the conditional
random earthquake function which is defined subsequently. The earthquake is input from the bottom
horizontally with the propagation velocity 250 m/s and
the duration is 20 s. The joint elements adopt the same
form with the pipeline element, while differentYoungs
module and density are defined.
For the deterministic analysis of the seismic
response of buried pipelines, the finite element method
is adopted to study the seismic response of buried
pipelines and the surrounding soil. The soil that
surrounds the pipeline is regarded as a solid-liquid
two-phase medium. The effective stress method and
nonlinear constitutive model of soil are used to study
the increase and the dissipation of pore water pressure during the seismic process. At the same time,
the contact interface between the pipeline and the surrounding soil is also included. Detailed techniques and
the dynamic parameters of the sandy soil and the contact surface can be referred to the related paper of the
authors (Ai & Li 2004).
4.2 Stochastic seismic analysis of underground

pipeline

model and real records for Soil Type IV.
In this paper, the random inputs and the uncertain soil

parameter are both considered and defined according to the reference (Li and Ai 2006): associated with
the seismic risk analysis, in the future definite time
range, the conditional random earthquake function is
defined as the case of the transcendental probability
of the earthquake motion being p = 3% and the peak
value of the seismic acceleration is 0.2 g; The internal
friction angle , which is defined as the random soil
parameter, has a logarithmic normal distribution with
the mean 150 and the standard deviation 0.30.
Under the proposed stochastic conditions, the probability density function (PDF) of the response for joint
E is presented in Figure 7, including the mean and the
standard deviation, typical instantaneous PDFs at certain instants of time, evolution of PDF against time
and contour to the PDF surface. These indicate the
503
stochastic fluctuation character of a nonlinear random

response.
CONCLUSIONS
A new approach is proposed for the stochastic seismic

response analysis of finite element modeled underground pipeline structures under earthquake excitations and uncertain soil parameters. The approach is
established based on the thoughts of the newly developed probability density evolution method. Associated
with the physics-based stochastic earthquake model
and the finite element method, the instantaneous probability density and the evolution process for the seismic response of underground pipeline can be studied
effectively. A computational example is investigated
with stochastic input and random soil parameters.
Some features of the responses are observed and discussed. It is found that the proposed method is of
high application in analyzing the stochastic response
of underground pipeline and the stochastic seismic
response for different structural forms requires further investigation. Furthermore, the seismic reliability
problem can be investigated easily by imposing the
failure criterion of the first passage problem reliability
theory.
ACKNOWLEDGEMENTS
The support of the Natural Science Foundation
of China for Innovative Research Groups (Grant
No.50321803) is gratefully appreciated.
REFERENCES
Figure 7. The PDF of the response for joint E.
Ai, X.Q. & Li, J. 2004. Seismic response analysis of underground pipelines using effective stress method. Proceeding of 13th World Conference on Earthquake Engineering.
Vancouver, Canada
Deodatis, G. & Shinozuka, M. 1988. Stochastic FEM analysis of non-linear dynamic problems. In: Shinozuka M
(ed). Stochastic Mechanics III, Department of Civil Engineering and Operations Research. Princeton University.
Iwan, W.D. & Huang, C.T. 1996. On the dynamic response
of non-linear systems with parameter uncertainty. International Journal of Non-Linear Mechanics 31(5):
63145.
Klosner, J.M., Haber, S.F. & Voltz, P. 1992. Response of nonlinear systems with parameter uncertainties. International
Journal of Non-Linear Mechanics 27(4): 54763.
Li, J. & Ai, X.Q. 2006. Study on Random Model of
Earthquake Ground Motion Based on Physical Process.
Earthquake Engineering and Engineering Vibration (in
Chinese) 26(5): 2126.
504
Li, J. & Chen, J.B. 2005. Dynamic response and reliability

analysis of structures with uncertain parameters. International Journal for Numerical Methods in Engineering
62(2): 289315
Liu, W.K., Belytschko, T. & Mani, A. 1986. Probability finite
elements for nonlinear structural dynamics. Computer
Methods in Applied Mechanics and Engineering 56:
6181.
Machida, H. & Yoshimura, S. 2002. Probabilistic fracture mechanics analysis of nuclear piping considering
variation of seismic loading. International Journal of

Pressure Vessels and Piping 79:193202
Nedjara, D., Hamanea, M., Bensafia, M., Elachachia, S.M. &
Breysse, D. 2007. Seismic response analysis of pipes by
a probabilistic approach. Soil Dynamics and Earthquake
Engineering 27:111115
Spanos, P.D. & Zeldin, B.A. 1998. Monte Carlo treatment
of random fields: a broad perspective. Applied Mechanics
Review 51(3): 21937.
505
Risk assessment for the safe grade of deep excavation

X.H. Bao & H.W. Huang
ABSTRACT: This paper embarks from risk idea, obtains safety factors of excavation and their percentages
from investigation results of excavation accidents, the safe grade of excavation is designed and analyzed. First, the
present developing situation, possibly existed risks in excavation projects and the state of present excavation grade
division in shanghai area is briefly introduced. Reasons of accidents are obtained from analyzing a large number
of excavation accidents. Later, fuzzy synthetic evaluation method is introduced and two levels of judgment
on excavation safety are carried on. Finally, the paper takes the example of shanghai international passenger
transport centre, the results conform to the actual situation.
INTRODUCTION
1.2 The present grade division and risk

factors in excavation
1.1 The present risk study of deep excavation

engineering
In recent years, excavation engineering develops
quickly, and excavations are becoming larger and
larger. The emergence of these large excavations has
brought new opportunities and challenges to engineering design, construction and management. On the
other hand, a great many of engineering practices
helped technical development, improved management
level, also made the risk analysis apply in many aspects
of engineering. Currently, most of the excavations are
mainly designed according to the specifications, rules
and accepted practice. However, excavations can be
designed according to the risk idea that is, to study
latent risk of excavation during the period of construction and operation, and reduce the unavoidable risk
to the insignificant level by increasing construction
cost. This method bases on the principles of reliability; it can make the probability of reaching limit
state small enough, and reduce the excavation risk
to acceptable level according to risk acceptance criteria. The research in this aspect at home and abroad is
still at a beginning stage, specialized research of risk
concerning deep excavation engineering is seldom,
the results obtained are almost qualitative, quantificational research is seldom, applicable risk analysis
and evaluation methods that are both qualitative and
quantificational are needed (Huang and Bian, 2005),
so it is difficult to construct probability models of risk
factors.
There are still some problems in present grade division

both at home and abroad. For example, uncertainty of
various factors which can influence the result, small
applicability scope of calculation theories and so on.
These will make construction cost be increased. Currently, the grade of excavation is divided according
to deformation values (Specification for Excavation
in Shanghai Metro Construction, SZ-08-2000). Every
grade has its control values. When the grade division is
not precise, if higher which means the safety warning
value is smaller, it will forecast the safe condition as
dangerous condition, then the excavation requests high
rigidity retaining structures, also more bracings and
bottom strengthening may be needed to resist deformation. Actually in this condition, investment has been
increased. If lower which means the safety warning
value is larger, the condition of excavation has already
been dangerous before forecasting, it may be too later
to remedy, so it can cause great loss and the work may
be delayed.
The characteristics of excavation engineering mean
that all owners have to face huge risk during construction. As the soil layer conditions and groundwater
circumstances are uncertain, technologies are complex, man may make mistakes in techniques and
managements, accidents may always happen in disadvantageous soil layers, so, there would not only exist
environment risk, but also other risks such as overspending and extension for completion date, etc. The
serious collapses appeared recently have revealed that
507
Table 1.
Reason
frequency
Percentage (%)
Investigation
Design
Construction
Supervision
Monitor
Owner
Total
16
125
176
5
10
12
344
4.65120
36.3372
51.1628
1.45350
2.90700
3.48840
100.000
Table 2.
tions.
2.2 Fuzzy synthetic evaluation method
Statistics of reasons for deep excavation accidents.
The fuzzy synthetic evaluation process (Liang and Bi,

2001) is as follows:
U is influence-factor set, U = {U1 , U2 , . . . Um },
Um (i = 1, 2, . . . m) is the i factor, in the secondary
synthetic evaluation, Ui = {Ui1 , Ui2 , . . . Uin }, Uin is
the sub-factor of Ui .
V is comment set, V = {V1 , V2 , . . . Vp }, Vj ( j = 1,
2, . . . p) is the j grade.
Construct evaluation matrix R
Retaining structures for the failed deep excava-
Retaining method
frequency
Percentage (%)
Row piles
Soil-nailing support
Deep mixing pile
Diaphragm
Excavated slope
Soil nailed wall
Others
Total
150
30
13
21
5
5
2
226
66.3717
13.2743
5.75220
9.29200
2.21240
2.21240
0.88500
100.000
Where R is evaluation matrix of single factor,

rij is the relative membership of Ui to Vj , Ri
(i = 1, 2, . . . m) is subordination vector.
A is weight set of main factors, A = (a1 , a2 . . . am ),
ai is the importance degree of Ui compared to the
other factors, 0 ai 1.
Choose composite operators and get composite
evaluation result B by multiplying A and R:
accidents can cause many big troubles in the excavation engineering. The excavation constructed in cities
can cause life and property loss to the third. The social
problems and public protest caused by excavation can
extend construction period (International Tunneling
Association, Second Unit, Guiding Rules of Tunnel
Risk Processing, 2002), so special attention should be
paid.
Where B is subordination vector of main factor to

comment congregation, R is evaluation matrix of
single factor, bi subordination vector.
Analyze results
2.3 Influence factors

2
2.1
RISK ASSESSMENT FOR THE SAFE GRADE

OF DEEP EXCAVATION
Reasons for deep excavation accidents
There are numerous reasons for deep excavation accidents, and a lot of scholars have papers in this aspect.
The following statistics are gotten from about 300 deep
excavation accidents (h > 6 m).
From table 1 it can be seen that construction and
Design are the main reasons. On the other hand, consider 226 accidents of these failed excavations; divide
them according to retaining method in table 2. Most
of them are retained by row piles. Although there may
be problems in this division, and it can not be said that
row pile is more damageable than other kinds of retaining structures, but accidents happened frequently, so
more attention should be paid on excavations retained
by row piles.
Considering investigations and analysis of other

researches (Yang and Ding, 1998), this paper considers five main factors that have influence on excavation
safety: size of foundation pit U1 , hydrogeology U2 ,
design U3 , construction U4 and surrounding environment U5 . Size of foundation pit includes area U11 ,
shape U12 , depth U12 . Hydrogeology includes layer
condition U21 , confined water U22 , drifting sand U23 .
Design includes calculation method U31 , value of
parameter U32 , value of load U33 , material selection
U34 and retaining structure U35 . Construction includes
if according to design and rules U41 , construction
method U42 , construction experience U43 , dewatering U44 , supporting U45 . Surrounding environment
includes dynamic load U51 , surcharge load U52 , adjacent construction U53 . In fuzzy synthetic evaluation,
these factors are also called indexes; the membership
between factors and their sub-factors (indexes system)
is shown in figure 1.
508
Table 4. The meaning of 19 scale.
Area
Shape
Size
Depth
Layer condition
Hydrogeology
Confined water
Drifting sand
Scale
Meaning(compare two factors)
1
2
3
4
5
The important degree is equal

The former is slightly important
The former is obviously important
The former is mightily important
The former is extremely important
* 2, 4, 6, 8 mean the middle value of two scales close together.
Calculation method
Safety factors of excavation
Parameter value
Table 5.
Judgment matrix U1 .
Load value
Design
Material selection
Retaining structure
According to
design and rules
U1
U11
U12
U13
U11
U12
U13
1
1/3
3
3
1
5
1/3
1/5
1
U2 , U3 , U4 , U5 can be gotten in the same way
Method
Construction
Experience
2.4.2 Construct judgment matrix

Judgment matrices can be constructed using Expert
Grades method, according to T.L. Sattys 19 scale.
Uij is subordinated to factor Ui , compare mutually the importance degree of every sub-factor Uij ,
ratio scale uij can be gotten, it reflects the relative
importance of two sub-factors, if the first sub-factor
compares to the second sub-factor and the result is uij ,
then the second compares to the first and the result
is uij = 1/uij , so judgment matrix U1 = (uij )nn can be
gotten as follows:
Dewatering
Supporting
Dynamic load
Surrounding
Environment
Surcharge load
Adjacent construction
Figure 1. Factors influencing the safe grade of deep excavation.
2.4.3 Weights of sub-factors

As it does not request high accuracy, to be simple,
calculate the maximum Eigen value max of the matrix
Ui and the feature vector A, and A is also the weight set
of sub-factors. The calculation process of A and max
is as follows:
Elements in the matrix multiply each other in rows
and open a n (n = 3) power to get A1i
Table 3. Weights of Ui .
Number
Influence factors
Weights
1
2
3
4
5
Size
Hydrogeology
Design
Construction
Surrounding environment
0.038
0.046
0.360
0.510
0.046
2.4 Weights of factors

2.4.1 Weights of the factors Ui
From the investigation of 344 excavation accidents
(h > 6 m), the following results can be gotten as shown
in table 3.
Use the Analytic Hierarchy Process. First, construct
a judgment matrix to calculate the maximum Eigen
value of the matrix and vector feature, and then test
the consistency of judgment matrix to determine the
weights of sub-factors (Wang and Huang, 2005).
So, A11 = 1
Normalize A1i to get A1i :
So, A11 = 0.296

Using the same method, A12 , A13 can be gotten, and
A1 = (0.296 0.120 0.584), A2 = (0.120 0.584 0.296),
509
Comments set
Table 6. Average random consistency index R.I .

Order of matrix
R.I .
0.52
0.89
1.12
A3 = (0.104 0.246 0.104 0.045 0.501), A4 = (0.203

0.086 0.466 0.203 0.042), A5 = (0.143 0.714 0.143).
The largest Eigen value calculation:
So, 1 max = 3.072, 2 max = 3.072, 3 max = 5.125,

4 max = 5.125, 5 max = 3.001
2.4.4 Test of consistency
On account of fuzziness of many factors and peoples
understanding is different because of subjectivity, so,
to overcome these subjective errors, the consistency
of judgment matrix should be tested. When stochastic
CR satisfies: C.R. 0.1, through Saatys CR standard,
the direction and extent of the improvement would be
controlled, and the optimal matrix with acceptable consistency could be available, that had preserved furthest
the decision makers primitive judgment information.
Consistency index C.I .
So, C.I .1 = 0.036, C.I .2 = 0.036, C.I .3 = 0.031,

C.I .4 = 0.031, C.I .5 = 0.0005
Check table 6 (Gong and Xu, 1986) to get the
corresponded average random consistency index R.I .
The average random consistency indexes R.I . of
every matrix are: 0.52 0.52 1.12 1.12 0.52
Consistency ratio C.R.
So, C.R.1 = 0.069 < 0.1, C.R.2 = 0.069 < 0.1,

C.R.3 = 0.028 < 0.1, C.R.4 = 0.028 < 0.1, C.R.5 =
0.001 < 0.1, the consistency of judgment matrices is
acceptable.
2.5 Comprehensive evaluation and results analysis
Fuzzy synthetic evaluation method is a kind of very
valid method that can completely evaluate the things
which are influenced by various factors, so, it can
be used largely in many projects and other systems.
This paper carries on two levels of judgment to the
excavation safety, first primary fuzzy synthetic evaluation of sub-factors, and then secondary fuzzy synthetic
evaluation of main factors.
Generally, the deep excavation safety can be divided

into 4 grades: very safe, safe, less safe, unsafe, these
can reflect the safety condition of deep excavation,
so the comments set V = {very safe, safe, less safe,
unsafe}. To calculate simply, quantify the comments
set, then V = {4 3 2 1}.
Construct evaluation matrix R
Obtain the relative membership degree of 19 factors
to comment set V and evaluation matrix R by using
Delphi Method. Invite 10 experts who know well about
the excavation as an evaluation group, and then get
marks of all indexes from experts. The steps are as
follows:
Invite 10 experts to evaluate, the factor sets
are U1 = {U11 , U12 , U13 }; U2 = {U21 , U22 , U23 }; U3 =
{U31 , U32 , U33 , U34 , U35 }; U4 = {U41 , U42 , U43 , U44 ,
U45 }; U5 = {U51 , U52 , U53 }. The relative membership
degree is rij : 0 rij 1, the size of rij shows the influence degree of the factor on safe grade of deep
excavation, and if bigger, more close to the corresponding grade, conversely, has little influence on the
corresponding grade. For the index Uij , if from 10
experts, mi experts consider the comment is v1 , ni
experts consider the comment is v2 , pi experts consider
the comment is v3 , qi experts consider the comment is
v4 , then the subordination vector Rij of the index Uij is:
Rij = (ri1 ri2 ri3 ri4 ) = (mi1 /10 ni2 /10 pi3 /10
qi4 /10), i = 1, 2, 3, 4, 5.
Evaluation matrices for the five main factors can be constructed by using subordination vectors as rows. So, R1 = (R11 R12 R13 )T ,
R2 = (R21 R22 R23 )T , R3 = (R31 R32 R33 R34 R35 )T ,
R4 = (R41 R42 R43 R44 R45 )T , R5 = (R51 R52 R53 )T .
Primary and secondary fuzzy synthetic evaluation
The paper uses weighted averaging fuzzy synthetic
evaluation model which can consider all factors and
single-factor evaluation results.
For the primary fuzzy synthetic evaluation, the
weighted averaging fuzzy synthetic model is:
Where A is the weight vectors of sub-factors, R is

single-factor evaluation matrix and R subordination
vector of sub-factor to comment congregation.
So, for every single-factor: Ri = Ai Ri (i=1,2,3,4,5 )
For the secondary fuzzy synthetic evaluation:
Where A is weight set of main factors, R is the

secondary evaluation matrix that is made up of Rj ,
B is subordination vector of main factor to comment
congregation.
So, for each single-factor: Bi = Ai Ri (i=1,2,3,4,5 )
510
The final value of comprehensive evaluation
Table 7. The safety grade of each main factor.
For the grade of deep excavation, we use the comments set V = {very safe, safe, less safe, unsafe}, after
qualification, V = {4 3 2 1}. W is the evaluation
value of deep excavation grade.
Factor
Value
Grade
Size
Hydrogeology
Design
Construction
Surrounding
environment
1.5 < W1 = 2.375 < 2.5

1.5 < W2 = 1.978 < 2.5
2.5 < W3 = 2.893 < 3.5
2.5 < W4 = 2.533 < 3.5
2.5 < W5 = 3.158 < 3.5
less safe
less safe
safe
safe
safe
So, for every single-factor: Wi = Bi V (i=1,2,3,4,5 )

If,
3.5<W <4.0, very safe;
2.5<W <3.5, safe;
1.5<W <2.5, less safe;
0.0<W <1.5, unsafe.
3.2.3 Secondary evaluation

The comprehensive evaluation matrix is:
3 THE EXAMPLE
3.1 The general situation of the project
Shanghai International Passenger Transport Centre
(west area) locates in south of east Daming Road,
west of Liyang Road, east of Gaoyang Road, north
of Huangpu River. The project covers an area of
128400 square meters, the shape of foundation pit is
rectangular and the depth is 13.10 meters.
3.2
As the weights for five main factors are known,

A = (0.038 0.046 0.360 0.510 0.046), so, the subordination vector B with assessment target U to
comment set V is:
Primary fuzzy synthetic evaluation
The membership of the factors and their sub-factors is

known in fig. 1.
3.2.1 Construct judgment matrix
Use Delphi Method. From 10 experts, for the index
U11 : the area of this foundation pit, no expert considers it very safe for the project, 3 experts consider
it safe, 6 experts consider it less safe, and 1 expert
considers it unsafe. So, the comment score to different safe grades are: 0, 3, 6, 1, so the subordination vector R11 = (0/10 3/10 6/10 1/10) = (0 0.3
0.6 0.1). After getting the other subordination vectors,
the evaluation matrix for U1 is as follows:
3.2.2 Primary evaluation

As is known, the weights vectors of the sub-factors
to main factor U1 is: A1 = (0.296 0.120 0.584),
B1 = A1 R1 = (0 0.475 0.425 0.1)
So, the evaluation value for single-factor U1 is:
W1 = B1 V = 2.375, in the same way, W2 = 1.978,
W3 = 2.893, W4 = 2.533, W5 = 3.158.
3.2.4 Evaluation results analysis

For the five main influence factors of the deep foundation pit, the safety grade division is in table 7:
Final comprehensive evaluation value:
So on the whole, the situation of this deep excavation is safe; the system safety can be basically
accepted. But according to the evaluation values for
single-factors, the size and hydrogeology are disadvantageous for the whole stability. To prevent this
deep excavation from accident, some measurements
should be taken during construction, for example
proper retaining method, good dewatering measure,
timely monitoring, etc.
The size and hydrogeology evaluation values can be
good advice to managers when they make decision. As
the size and hydrogeology are disadvantageous for this
excavation safety, so they are considered to be two risk
factors, then during the project management process,
risk management can be carried on to decrease the
probability of failure that may be caused by these two
risk factors. Through the master and supervision of
risk factors, managers can adopt active measures and
scientific management.
511
REFERENCES
CONCLUSION
In deep excavation engineering, using fuzzy synthetic

evaluation method can quantify qualitative analysis
and make inaccurate expression become numeral,
so the evaluation process is more scientific. The
safety influence factors in deep excavation engineering are various; the paper obtains the safe factors and
its weights from the investigation results of excavation accidents, so the safe grade evaluation is more
scientific and rational.
In the primary evaluation, the results are advantageous for discovering the trouble and weakness that
lurked in engineering; they can provide decision to
managers in order to do prevention, so the whole safety
level of deep excavation is raised.
The example demonstrates that the results conform
to the projects actual situation; and they can help
raising the overall safety level of the excavation and
provide some reference values for the projects. Thus
this paper has proved that the Fuzzy Synthetic Evaluation, Analytic Hierarchy Process and Delphi Method
are favorable in deep excavation engineering.
Guo, Z.W. 1986. Risk Analysis and Decision. Beijing: China

Machine Press.
Huang, H.W. & Bian, Y.H. 2005. Risk management in deep
excavation construction. Chinese Journal of Underground
Space and Engineering 1(4): 611614.
Liang, S. & Bi, J.H. 2001. Fuzzy synthetic evaluation method
of construction engineering quality grade. Journal of
Tianjin University 34(5): 664669.
Shanghai Metro Construction Corporation LTD, SZ-082000. Specification for Excavation in Shanghai Metro
Construction. Shanghai.
Soren, D.E., Per, T. & Jorgen, K. 2004. Guidelines for tunneling risk management: International Tunneling Association, Working Group No. 2. Tunnel and Underground
Space Technology 19(2004): 217237.
Wang, Y. & Huang, H.W. 2005. Hierarchyfuzzy synthetic
safety evaluation method of tunnel in subway. Journal of
Underground Space 24(3): 301305.
Yang, Z.M. & Ding, X.G. 1998. Safety accident and reason
analysis of excavation engineering. Safety Regulations for
Geological Prospecting Operation 5(1): 1013.
ACKNOWLEDGEMENT
The author is very grateful of Doctor Y.H. Bian for the
investigation results of deep excavation accidents.
512
Multi-factors durability evaluation in subway concrete structure

C. Chen & L. Yang
C. Han
School of Economics & Management, Tongji University, Shanghai, P.R. China
ABSTRACT: In this paper, an erosion model of subway concrete lining structure was proposed, then factors
which the influence durability of subway structure were determined, and the framework of evaluating measure
index was established. Finally, Fuzzy Multiple Attribute Decision Making was used to carry out durability
prediction model on the basis of the single factor evaluation in the subway structure. Through case study, the
durability of subway structure is evaluated by the methods in this article and get good or bad schemes.
2 ANALYSIS OF SUBWAY STRUCTURE

DURABILITY
INTRODUCTION
Shanghai Subway belongs to the Fast OrbitTransportation System in the city. Because of such characteristics
as being free from weather influence and being safe
and fast, it becomes the key vehicle within network of
the passenger transportation (Shen et al. 1998). The
subway structure is a permanent concrete structure, so
the investment is usually huge, and it always crosses
key districts of city. The requirement of durability of
subway is higher than that of other engineering (the
base period for design is 100 years), and the problem
of durability is of most urgency.
The factors influencing concrete structural durability are numerous and very complicated, and they
are interrelated and influence on each other. Therefore, it has two kinds of manifestations: one is the
uncertainty of the occurrence of an affair, and the
other is the uncertainty of the behavior of the affair.
The existing evaluation methods of concrete durability mostly evaluate the influence of a single factor by
setting up a model. However, actually, the concrete
structure durability is a comprehensive function containing many factors, and its evaluation is a compound
decision process of multiple attributes, multiple factors and multiple indexes. That is to say, the evaluation
of concrete durability is a FMADM (Fuzzy Multiple
Attribute Decision Making) (Li 2002, Leung & Cao
2000) problem. In view of this problem, FMADM is to
be used to carry out the multiple factors evaluation on
the basis of the single factor evaluation in the subway
structure (Heilpern 1997, Zimmermann 1987).
2.1
Some concepts
The durability of concrete structure is the capability

of the structure which keeps its safety, function and
requirement of appearance under suitably maintaining
condition during the base period of design where the
effect of the environment must be considered. Here,
the environment is a general concept, which contains
any factor resulting in degradation and destruction of
the performance of the structure.
Therefore, the durability of concrete structure is a
integrated functional index which is concerned with
not only raw materials and proportioning of concrete
but also mechanics environment, natural environment,
service environment and production technology of
concrete (Tan 2003). When studying the durability
of concrete structure, two states, which include the
limit of bearing capacity and the limit of function
and appearance requirement, should be considered.
Namely, it is better to obtain economical and rational life-span on the premise of ensuring the structure
safety.
2.2
Influencing-factors of subway structure

durability
There are a few factors which can modify the durability of concrete structure. When proportioning has been
confirmed, the maximal modifying factor to durability is its environment. Here, the factors of environment
513
mainly include carbonation, chloride ions ingress, sulfate attack and cycle of freezing and thawing (Du &
Zhang 2003).
The erosion model of concrete lining structure is
shown as Figure 1.
The concrete lining of subway structure is subjected to interactive action of multi-destroy factors that
maybe come from interior environment of atmosphere
or form exterior corrosion of geotechnical medium.
Obviously, durability of subway structure should be
analyzed on the basis of its circumstance as it is.
Compared with others concrete structure, the subway structure has two prominent characters. Firstly,
the temperature of subway structure changes very little because it is long-term located under the surface
of earth. Therefore, the probability of destruction of
subway structure is very little owing to the action of
cycle of freezing and thawing. Secondly, the subway
is operated by the traction of electric power. As a

result of this, it produces stray electrical current which
plays a very important role in causing the corrosive
destruction in subway structure.
Therefore, the influencing-factors of durability of
subway structure mainly include stray current corrosion, chloride ions ingress, sulfate attack and carbonation. The life-span of durability of structure is a
function of these influencing-factors.
2.3 Evaluating measure index framework of
subway structure durability
As described previously, the influencing-factors of
durability of subway structure mainly include stray
current corrosion, chloride ions ingress, sulfate attack
and carbonation. Actually, each factor interact with
many others relative factors. Due to lack of space,
this paper analyzes only these four factors described
previously.
Table 1 shows the evaluation measure index framework of durability of subway structure.
2.4 Evaluation of subway structure durability
Durability of subway structure is a comprehensive
function containing many factors, such as stray current corrosion, chloride ions ingress, sulfate attack and
carbonation (Figure 1). At the same time, the single
factor is always the function of other influential factors
(Table 1).
Figure 1. Erosion model of concrete lining structure.
Table 1.
Evaluation measure index framework of durability of subway structure.
Objective level
Primary criteria level
Durability of
Subway
Structure
Stray Current
Corrosion
Secondary
criteria level
Coltage factor
Chloride ions
Ingress
Resistance
factor
Concrete
characteristic
Sulfate Attack
Diffusivity
Coefficient
Exterior factors
Interior factors
Carbonation
Exterior factors
Interior factors
Operational level
Load current of train; Contact potential difference;
Transformer substation spacing
Electrical resistivity of concrete; Transition resistance;
Longitudinal resistance of rail
Dosage and variety of cement; Variety and graded of mixture;
Admixture variety; Water cement radio; Air content;
Curing time; Hydration degree
Environmental temperature and humidity; Open-assembly
time; Chloride ion concentration; Stress state; Crack
Sulfate ion concentration; Magnesia ion; Chloride ion;
PH value; Alternation of wetting and drying; Freezing
and thawing cycle; Stress state; Crack
Dosage and variety of cement; Variety and graded of
mixture; Water cement radio; Admixture variety; Pore
content and distribution
Light and temperature; Relative humidity; co2 concentration;
Stress state; Crack
Dosage and variety of cement; Variety and graded of
mixture; Water cement radio; Compressive strength
of concrete; Construction quality; Curing method
514
where Lcr is the durability of subway structure, 1 is

the influential factor of stray current corrosion, 2 is
the influential factor of chloride ions ingress, 3 is the
influential factor of sulfate attack, 4 is the influential
factor of carbonation, i is the influential factor index
of the single factor.
The durability evaluation of a single factor is related
to the same subject that influences the durability of
subway structure, and it is the basis of the durability evaluation containing multiple factors. However,
because of the different damage mechanism of each
single factor, so, each factors evaluation model and
evaluation method are different, and the influential
degree to durability of subway structure of each factor
is of difference.
For example: the sulfate attack is the entrance of
SO42 from external corrosion media into the inner
concrete through the surface of the concrete, and if
chemical reaction happens between SO42 and substance of the cement, such as Ca(OH)2 , then the
chemical reaction will generate Aft, which is of distensibility and of hazard to concrete by causing expansive
cracks, and the concrete is in fragile and loose state.
Therefore, the procedure of sulfate attack is from the
external to the interior, and the damage degree deepens with the corrosion depth. However, chloride ions
ingress through the concrete surface and the permeation is of little damage to concrete. But, the result of
the permeation changes the alkali environment of the
concrete, which makes the passivating film of concrete structure disappear to induce the rust-eaten of
bars, thus, the concrete structure is damaged. Therefore, the damage procedure of chloride ions ingress is
from the interior to the external. The damage degree is
slight in initial state, and develops in geometric series
with the time.
Durability of subway structure is the comprehensive
interaction of multiple factors, and it is insufficient to
analyze the issue from the action of the single factor or
the simple perspective of single factors. Because of the
fuzziness of many factors, it is not suitable to evaluate
it with an accurate mathematic model. Therefore, to
guarantee the objectivism of the evaluation of subway
structure, it is of ultimate caution to choose a suitable
evaluation method.
On the basis of the Fuzzy Multiple Attribute
Decision Making theory, the article sets up fuzzy
multi-factors evaluation model for durability of subway structure applying the method AHP and expert
investigation to determine the weight of each factor.
multi-factors and indefinite problems. In this paper,

the author use fuzzy multiple attribute decision model
to evaluate the durability of subway structure. The
steps are described as follows:
1. Suppose the collection of evaluating scheme
X = {X1 , X2 , X3 , , Xm }, which X are m dimensions pending optimum schemes. Determine the
index vector Gm = {G1 m , G2 m , , Gnm }T if each
scheme has n dimensions evaluating index.
2. Determine attribute value matrix of scheme collection Gij (i = 1, 2, ...n, j = 1, 2, , m), which is
abbreviated as r = (rij )nm .
3. Transform attribute value matrix to relative grade
of membership matrix.
Multi-targets decision-making is lack of commensurability. Namely, each target hasnt unified standard
of measurement and so they cant be compared each
other. For compared easily, the value of attribution
should be quantified and changed them all into [0, 1].
Based the type of target, different quantified method
should be chosen. The target can be divided the following types: the cost type (the smaller one is better
one), the benefit type (the larger one is better one),
the moderation type (appropriate one is better) and
interval type (be located in stable zone is better).
Here the benefit type is used as quality matrix of
target attribute of durability of subway structure. (The
life-span of durability is older, the effect is better).
where mk = minimum eigenvalue; Mk = maximum

eigenvalue; and dk = [mk , Mk ]
The base period for subway design is 100 years.
Hence, here one gets mk = 100. Whats more, considering economical efficiency and demand of sustainable
development in urban construction synthetically, here
one gets Mk = 150.
4. Ascertain the weight of each factor, the weight
vector of attribution is (w1 , w2 , , wn )T , here
n

wj = 1, and wj 0. The weight vector is ascerj=1
tained by the method of arrangement analysis,

which can be seen in 4.4 in details.
5. Calculate the equation:
FUZZY MULTIPLE ATTRIBUTE DECISION

MODEL (FMADM)
Fuzzy Multiple Attribute Decision Model (FAMDM)

is a comprehensive analysis method which can solve
Ai is the collection of evaluating result. aj (j =

1, 2, m) is the fuzzy evaluating value of each
515
Table 2. Alternative scheme of concrete proportioning and attribute of influencing-factors of durability.

Attribute of influencing-factors
of durability (years)
Primary material content of concrete (kg/m3 )

Scheme
Number
Cement
Water
D300
D310
D320
D331
D332
339
254
138
163
110
165
161
159
150
157
Sand
Cobble
(5 25 mm)
Fly
ash
Mineral
Powder
784
713
718
703
712
1148
1138
1051
1080
1062
0
108
0
110
110
0
0
228
110
148
HLC
Stray
Current
Corrosion
Chloride
ions
ingress
Sulfate
attack
Carbonation
45
45
45
45
45
114
126
134
143
180
110
159
134
150
173
163
144
119
122
135
112
100
150
113
118
D300: Primary Standard Concrete D310: Concrete Containing Fly Ash D320: Concrete Containing Mineral Powder
D331 D332: Concrete Containing Fly Ash and Mineral Powder
HLC: crack-resistance and seepage-proofing agent
scheme. The superior or inferior rank of schemes

is estimated by the evaluating value (01), and the
evaluating value is larger, the scheme is better.
4
4.3 The relative quality matrix of evaluation

system of subway durability
According to formula (2), the relative quality matrix of
evaluation system of subway durability is written as:
ILLUSTRATIVE EXAMPLE
Here are 5 Alternative schemes for researching high

performance concrete. Through the researching to a
series of experiments and correlative durability works,
the life-span of durability of these schemes is obtained
under the action of stray current corrosion, chloride
ions ingress, sulfate attack and carbonation.
The Alternative schemes of concrete proportioning
and the researching attribute of the life-span of durability under the action of each factor individually can
be shown in Table 2.
4.1 The collection of decision-making evaluation
schemes and collection of attributions
As analyzed previously, the collection of decisionmaking evaluation schemes can be described as:
X = {A, B, C, D, E}; the collection of attribution can
be described as:
G = {G1 , G2 , G3 , G4 }T = {Stray Current Corrosion,
Chloride Ions Ingress, Sulfate Attack, Carbonate}T
4.2 The evaluating attribution matrix of subway
durability
The value of attribution can be described as following
matrix:
4.4 The weight of each factor in the evaluation

scheme
The ascertainment of weight adopts to the method
AHP. According to the stray current corrosion and the
concentration of CO2 , analyze the influence of single
factor to the life-span of durability of concrete. And
use Delphi method to analyze and get the comparing
estimation matrix between any two ones in these factors. The standard of ascertained index is: when one
compares with another one, the index get 1 if both is
the same important; the index get 3 if the former one is
only a bit more important than the latter one; the index
get 5 if the former one is markedly more important
than the latter one; the index get 7 if the former one
is much more important than the latter one; the index
get 9 if the former one is absolutely more important
than the latter one. The calculation procedure in detail
can be shown in reference (Wei 2002). Here get the
solution of comparison:
516
The present results show that scheme E is the best

proportioning scheme of subway concrete which obviously enhances durable life-span (the durable life-span
is longer than 100 year) by mixing fly ash and mineral
powder reasonable. The recommendation uses the best
proportioning for the serial number E.
It should be mentioned that superimposed effect
between various factors is not to be considered in
the paper. For instance, carbonation affect chloride
ions ingress, chloride ions ingress affect sulfate attack.
Therefore, durability of subway structure considering superimposed effect between various factors needs
further research.
Use the method of square root to make certain the

weight of each index and get:
These weight values are available through proving.

4.5
Fuzzy evaluation of the system of

subway durability
Using formula (3), the fuzzy values are:
ACKNOWLEDGEMENT
Therefore, the alternatives Ai (i = 1, 2, 3, 4, 5) can
be ranked as A5 > A3 > A4 > A2 > A1 , the relevant
order of proportioning from good to bad is sorted as
followed: E > C > D > B > A
Obviously, the scheme E is best one. The recommendation uses the best proportioning for the serial
number E.
CONCLUSION
Durability of subway structure is a problem of comprehensive function under many factors, and its evaluation
is a compound decision process of multiple attributes,
multiple factors and multiple indexes. In view of
this problem, the paper determined influencing-factors
of durability of subway structure for stray current
corrosion, chloride ions ingress, sulfate attack and
carbonation, established the evaluating measure index
framework of durability of subway structure. Further,
FMADM is to be used to carry out the multiple factors
evaluation on the basis of single factor evaluations in
the subway structure. Then, the durability fuzzy values of alternative scheme of concrete proportioning
are obtained. Finally, these values are ordered by their
size and make them clear which scheme is good or bad.
This work was supported by the National Natural Science Foundation of China under Grant Nos.50678135.
The author would like to acknowledge Prof. Yang
Linde for this research project.
REFERENCES
Du, Y.J. & Zhang, H.H. 2003. Elementary study in evaluation of enduring span for HPC of subway engineering.
Northwest Water Resources & Water Engineering (3):
4953.
Heilpern, S. 1997. Representation and application of fuzzy
numbers. Fuzzy Sets and System: 259268.
Leung, L.C. & Cao, D. 2000. On consistency and ranking of alternatives in fuzzy AHP. European Journal of
Operational Research 124:102113.
Li, T.J. 2002. Theory and application of Fuzzy Multiple
Attribute Decision Making. Beijing: Science Press.
Shen, G. 1998. The NO. 1 subway project. The series of design
and construction of shanghai municipal works. Shanghai:
Shanghai Science and Technology Press: 13.
Tan, W.Z. 2003. Holistic view of durability of concrete
structure. Architecture Technology (1): 1922.
Wei, R.Q. 2002. Operational research. Beijing: Tsinghua
University Press,: 461466.
Zimmermann, H.J. 1987. Fuzzy Set Theory and its Application. Kluwer2Nijhoff Publication, USA,: 5152.
517
The use of artificial neural networks to predict ground movements caused by

tunneling
I. Chissolucombe, A.P. Assis & M.M. Farias
University of Brasilia, Brazil
ABSTRACT: During the design phase of a tunnel, one of the concerns of tunnel engineers is estimating ground
movements induced by tunneling, in order to take steps to prevent or minimize possible damage caused by these
movements, to adjacent structures. In this work, an estimate is made of displacements in the ground brought
on by tunnel excavation by using an artificial intelligence technique called Artificial Neural Networks (ANN).
The computer tool utilized was the MatLab program and the networks were of the feedforward type, with the
Resilient Backpropagation (trainrp) learning algorithm.
INTRODUCTION
Tunnelling provokes changes in the stress state of

a ground and consequently, some displacements are
generated that spread across the excavations zone of
influence, which may induce damage in the structures located therein. Several methodologies have been
utilized to estimate the displacements generated by
the excavation of a tunnel and the damage that such
displacements can cause to the structures lying adjacent to the excavation area. These methodologies are
distributed by the empirical, analytical and numerical approaches. The empirical approaches are easy
to use, but have the disadvantage of not considering the ground resistance and deformability parameters (Chissolucombe et al., 2005a). The analytical
approaches are also easy to use, but have the primary
disadvantage of determining the various coefficients
necessary for the use thereof. In using numerical
approaches, one must have access to a test program
that can supply the necessary parameters in a way
as to be able to use constitutive models that faithfully represent the conditions of the ground and often
haveas in the case of three-dimensional analysesa
high computational cost.
The aim of this study is to estimate the displacements of the ground induced by tunneling and the
inflection point of the surface settlements basin utilizing one of the artificial intelligence (AI) techniques,
called artificial neural networks. To verify which of
the input variables had the greatest impact on the
output variable of the neural network a sensitivity
analysis was carried out utilizing the values of the
weights between and among the network connections
as proposed by Garson (1991).
In the training of the neural network, two data sets

were used, one of which was obtained by instrumentation during the excavation of the Metr-DF (subway) in
the city of Brasilia, and the other was obtained through
numerical simulations by the Finite Element Method.
In all of the situations, the training algorithm utilized
was Resilient Backpropagation (trainrp), with sigmoid
transfer functions.
The use of two sets of data obtained differently is
due to the following factors: (a) the input variables of
the set of data of Metr-DF were mostly geometrical
parameters of the tunnel or of the ground , having the
SPT value as a single parameter of resistance; (b) need
to know the behavior of the neural network upon the
presentation of the input variables as cohesion and the
angle of friction, values obtained through numerical
simulations.
2 ARTIFICIAL NEURAL NETWORKS
A neural network is a massively parallel distributed
processor that has a natural propensity for storing
experiential knowledge and makes it available for use.
It resembles the human brain in two respects: (a)
knowledge is acquired by the network from its environment through a process of learning, and (b) connection
strengths among neurons known as synaptic weights
are used for storing the acquired knowledge. (Haykin,
2001).
The most important property of Artificial Neural
Networks is their ability to learn from their environment and to improve their performance with such
learning. This learning occurs when the neural network
reaches a generalized solution for a particular class
519
Figure 2. The geological profile of the tunnel from the South

Wing.
Table 1.
Figure 1. Types of partialization of the cross section.
Input statistics for data sets.
Variables
of problems. In 1970, Mendel & McClaren (cited by

Haykin, 2001) defined learning as being an interactive
process whereby the free parameters of a neural network are adapted through a process of stimulation by
the environment in which the network is inserted. This
definition implies the following sequence of steps: (a)
the neural network is stimulated by the environment;
(b) the neural network undergoes modifications in its
free parameters as a result of this stimulation; and (c)
the neural network responds to the environment in a
new way, due to the modifications that occurred in its
internal structure.
The set of well-defined rules for the solution of a
learning problem is called a learning algorithm. There
are various kinds of learning algorithms, and the main
difference among them is the way in which the weights
are adjusted.
3
3.1
MODELING WITH ARTIFICIAL NEURAL

NETWORKS
Modeling 1 (surface marks)
In this modeling, the database utilized was comprised

of data obtained from the instrumented sections during
the excavation of the tunnel for Metr-DF (subway) in
the city of Brasilia.
The Brasilia Metro has an extension of 42 km, being
12 km underground. The South Wing section is totally
underground, with 7.2 km of length, and encompasses
nine stations. These stations were built by the cutand-cover method and the tunnel excavated by the
sequential excavation method, following the NATM
(New Austrian Tunnelling Method) principles. The
geometry shape of the tunnel cross-section was an
ovoid, with an equivalent diameter of 9.6 m. The types
of partialization of the cross section of the Metr DF
tunnel are represented in Figure 1. The South Wing
tunnel of the Brasilia Metro was mostly excavated
inside a geological domain, called Paranoa group,
which encompasses layers of porous clay and residual
H (m)
D (m)
SPT roof
SPT side
SPT floor
Training
Testing
Training
Testing
Training
Testing
Training
Testing
Training
Testing
Mean
Minimum
Maximum
10.24
10.03
8.49
8.61
22.92
22.70
3.75
3.75
8.72
8.79
6.70
7.70
7.26
7.26
0.40
0.90
1.50
1.50
3.50
3.50
14.30
12.90
9.56
9.46
50.00
50.00
8.50
8.50
18.00
17.00
soil of slate (Chissolucombe et al., 2005b). The geological profile of the tunnel from the South Wing is
represented in Figure 2.
The input variables in the network were as follows:
ground cover above the tunnel roof (H), the type of
partialization of the section (TS), the equivalent diameter of the section (D), the length of the balance (x),
the level of the water table (N.A.), the SPT value of
the tunnel roof, sides and floor. The output variables
were the values of the stabilized surface settlement on
the tunnel axis obtained from adjustments after taking
readings of the surface marks and the inflection point
(i). The medium, minimum and maximum values of
some input variables are presented in Table 1.
The set of data was comprised of 74 pairs of examples, 62 being utilized in the training phase and 12
in the testing phase. The training was performed considering each output separately; this procedure was
adopted because the performance of the network was
better than when both outputs were considered simultaneously. The architecture that offered best generalization was 8-40-1 for the output variable maximum
surface settlement and 8-10-1 for the output variable
inflection point of the settlement basin. Figures 3, 4,
5 and 6 show the results obtained during the training
and testing phases of the neural network, and an average of 3,000 interactions was necessary to obtain the
performance presented.
520
Figure 3. Comparison between the predicted and measured results for the maximum surface settlement during the
training phase.
Figure 5. Comparison between the predicted and measured

results for the inflection point of the settlement basin during
the training phase.

results for the maximum surface settlement during the testing
phase.
3.2
Modeling 2 (numerical simulation)
In this modeling, the data obtained from twodimensional numerical simulations under conditions
of flat deformation was utilized, with the Plaxis finite
element program (Brinkgreve &Vermeer, 1998) developed at Delft University of Technology. The section to
be excavated was partialized and the numerical simulation obeyed the following steps: (a) generation of the
initial stress based on coefficient of lateral earth pressure (Ko); (b) head excavation with application of 50%
of the nodal forces in the area around the excavation;
and (c) activation of the support with application of
the remainder of the load. The excavation of the sides,
nucleus and inverted arch followed the same sequence
described above. The finite element mesh was comprised of 275 triangular elements with 6 nodes each,
totaling 636 nodes and 825 Gauss points.

results for the inflection point of the settlement basin during
the testing phase.
The model constituent adopted was the MohrCoulomb, the main parameters of which are cohesion
(c), angle of friction (), Young module (E), and Poisson coefficient (). For the modeling, 432 numerical
simulations were performed, varying the cohesion values (10; 20; 30; 40; 60; and 80 kPa), angle of friction
(17 ; 22 ; 26 and 30 ),Young module (3; 6; 10; 15; 20
and 30 MPa) and coefficient of lateral earth pressure
(Ko 0.35; 0.55 and 0.60). The value of the diameter (D) of the soil cover above the tunnel roof and of
the Poisson coefficient were constant (4.5 m, 10 m and
0.33 respectively).
521
Figure 7. Performance of the neural network in the training

phase for the output variable maximum surface settlement.
Figure 9. Performance of the neural network in the training

phase for the output variable settlement at tunnel roof.
Figure 8. Performance of the neural network in the testing

phase for the output variable maximum surface settlement.
Figure 10. Performance of the neural network in the testing

phase for the output variable settlement at tunnel roof.
The following variables: cohesion, angle of friction, Young module, and coefficient of lateral earth
pressure, constituted the input variables in the neural network, while the maximum surface settlement
and the settlement at tunnel roof constituted the output variables in the neural network. Out of the total
of 432 pairs of examples available, 352 were utilized
in the training phase and 80 were used in the testing phase. The best network architecture was 4-40-1,
and each output variable was calculated separately.
An average of 1,500 interactions was necessary for
training the neural network in order to obtain a good
generalization.
Figures 7, 8, 9 and 10 show the results obtained in

the training and testing phases for the output variables
maximum surface settlement and settlement at tunnel roof with respective correlation coefficients (R)
of the best linear adjustment. According to Figures 7,
8, 9 and 10, excellent performance was obtained for
both variables in the training and testing phases.
SENSITIVITY ANALYSIS
To verify the relative importance of each input variable

in the estimate of the output variables, a sensitivity
522
Figure 11. Relative importance of the input variables in

the network to estimate maximum surface settlement and
inflection point (Modeling 1).
Figure 12. Relative importance of the input variables in the

network to estimate the maximum surface settlement and the
settlement at tunnel roof (Modeling 2).
analysis was conducted according to the that proposed

by Garson (1991). Figure 11 shows the sensitivity
analysis for Modeling 1, i.e., for the case where the
network was trained using data obtained by instrumentation during the excavation of the tunnel for the
Metr-DF (subway) in Brasilia. For the output variable
maximum surface settlement, the SPT resistance
parameter was the one that presented the greatest relative importance and its maximum value was 32.26%
occurring for the SPT in the tunnel side; the input variable balance was the least significant, with a relative
importance of 3.84%. In case of the output variable
inflection point of the settlement basin, the most significant input was also the SPT in the tunnel side, with
a value of 17.52%, and the least significant input was
balance, 7.13%.
In Modeling 2, where the data were obtained
through the numerical simulations with the Plaxis
finite element program and the output variables were
maximum surface settlement and settlement at tunnel roof, the relative importance of the input variables
was nearly identical to the two output variables. In that
case, the resistance parameters were the ones that presented greater relative importance, having a cohesion
value of 48%, followed by a Young module of 28%,
friction angle of 15%, and coefficient of lateral earth
pressure with an average value of 9% (Figure 12).
CONCLUSIONS
The following conclusions were reached after the training of the two sets of data utilizing the AI technique
called artificial neural networks: (a) the computational
tool utilized proved highly effective in the estimate
of the displacements induced by tunneling and of the
inflection point of the surface pressure basin, having
obtained high correlation coefficient values and very
minor errors when comparing the values measured
with the values estimated by the network; (b) when the
objective is to excavate a tunnel in a region where a set

of data on previous excavations is available, the use of
neural networks will provoke a drastic reduction in the
programs of geotechnical geological investigation and
measurements of the construction work; (c) the network trained with the data obtained by instrumentation
during the excavation of the Metr-DF tunnel exhibited an inferior performance to the network trained
with the data obtained through numerical simulations.
This is due to the noise that the data obtained in the construction work presents. Nevertheless the performance
of this network was reasonable with average correlation coefficients of 0.99 for training and 0.95 in the
test phase; (d) the neural networks present an advantage in relation to traditional empirical methods as
they permit the consideration of variables such as: constructive method, water level, geometry of the tunnel,
and geometrical parameters and parameters of resistance of the ground ; (e) compared with the numerical
methods it has the advantage of estimating with high
precision the displacements induced by the excavation, precision that is not always achieved in numerical
analyses due to factors like: lack of an adequate constitutive model; difficulty in representation of the real
geometry; consideration of the tridimensional effect
in bidimensional analyses; (f) the sensitivity analysis
showed in the two modelings that the parameters of
resistance and deformability of the ground have the
greatest impact on the estimate of displacements, with
the SPT values on the sides, roof and floor of the tunnel
in modeling 1 and soil cohesion and Youngs module
in modeling 2; (g) the main disadvantage of neural
networks is the lack of a good set of data for training.
REFERENCES
Brinkgreve, R.B.J. & Vermeer, P.A. 1998. Finite Element
Code for Soil and Rock Analyses. Plaxis Manual. Rotterdam, Netherlands.
523
Chissolucombe, I., Assis, A.P. & Farias, M.M. 2005a. Mtodos para previso dos deslocamentos do macio induzidos
por escavaes subterrneas. 5 Simpsio Brasileiro de
Aplicaes de Informtica Em Geotecnia (INFOGEO).
Minas Gerais.
Chissolucombe, I., Assis, A.P. & Farias, M.M. 2005b. Soilstructure interaction and its influence on displacements
induced by tunnel excavations. Fifth International Symposium Geotechnical Aspects of Underground Construction in Soft Ground. Netherlands.
Garson, G.D. 1991. Interpreting neural-network connection
weigts. AI Expert, 6(7): 4751.
Haykin, S. 2001. Redes Neurais: Princpios e Prtica.
Bookman, 2o edio, RS, 900p.
524
Research and application of road tunnel structural optimization

W.Q. Ding & Y. Xu
Department of Geotecnical Engineering, Tongji University, Shanghai, P.R. China
ABSTRACT: The optimization process of road tunnel is established in this paper with an engineering example.
Through numerical computation, we can find a serious stress concentration occurring at the spring of arch.
To solve this problem, a tunnel section shape optimization method is established at first. In this method, the
dimensions of the tunnel section are taken as the design variables, the stress of spring is taken as the objective
function. No augmentation of the excavated area and the satisfaction of construct boundary are taken as the
constraint condition. A Complex Method is used as the optimization method and C++ language is used to
program and implement the optimization. Using this optimization method, the stress condition of the tunnel
lining can be greatly improved. Then, the optimization method for the thickness and reinforcement of the tunnel
lining is also established and the construction cost is reduced.
INTRODUCTION
With the rapid development of highway in mountain

area, the quantity of road tunnel becomes more and
more. However, the cost of road tunnel is higher than
that of road or bridge, so, how to reduce the total cost
of the road tunnel, and how to construct road tunnel in
the most economic way are the imperative problems
which need to be solved by designers.
Optimization design of the road tunnel is an effective method to reduce the construction cost. But now,
in the optimization design of the road tunnel, the traditional passive analysis method is used most. The
designers often define several feasible cases, then,
choose one from them through computation and analysis. The disadvantage of the traditional passive analysis
method lies in that it can not contain all conditions, and
it is time consuming. And the numerical optimization
method is not used widely, more research is needed
and imperative.
The numerical optimization method is a new branch
of applied mathematics. In recent twenty years, it
develops very quickly with the widespread application of computer. Applying the numerical optimization
method in civil construction design makes the traditional passive analysis develop to the active search.
This is a grate progress in the civil engineering design,
and the numerical optimization method is a more scientific, more effective and more economic method in
structural optimization design.
But in the domain of optimization of tunnel and

underground structure, there are many problems. Due
to the complexity of tunnel and underground structure, a great many of parameters should be optimized
(such as parameters of surrounding rock, parameters of
support structure and decision options in constructing
and so on). As to the optimization, there are problems
such as multi-variable and highly non-linear, so the
research and its application are still in the start stage,
the references and examples are few.
With an engineering example, an optimization
design method of automatic search is expounded in
this paper. The section shape and the lining structure
of the tunnel can be optimized. A Complex Method is
used as the numerical optimization method, the C++
language is used to implement it. With the optimization analysis and design, the stress condition of the
structure is greatly improved, and the cost of the tunnel is effectively reduced. This is of great practical
significance to the construction of road tunnel.
2 THE THEORY OF COMPLEX METHOD

2.1 The mathematic model of complex method
The mathematic model of complex method usually can
be expressed as follows:
Solve: design variable X
Min
f (x)
525
S.T (Subject to constraint condition):

Inequation constraint condition:
Boundary constraint condition:
where ai is the upper limit of the variable; bi is the

lower limit of the variable.
2.2 The iterative process of complex method
2.2.1 The building of initial complex
The complex is composed of k(k > n + 1) vertexes.
There are two methods to define these vertexes.
a. Definitive method
Designers can define these vertexes by themselves
according to the properties of the problem.
b. Random method
The vertexes are defined by the following formula:
where i is the number of variable; j is the number of

vertex; ij is a random value between 0 and 1.
The vertexes defined by random method conform
to all boundary conditions, but do not always conform
to all inequation of constraint conditions.
Supposing there are s(1 s k) vertexes conform
to all constraint conditions, the center X s of the valid
vertexes can be defined by following formula:
The n-s vertexes that can not conform to all constraint

conditions can be dealt with by following formula:
If the new vertexes still can not conform to all constraint conditions, they are dealt with formula (5) again
until they can conform to all constraint conditions.
2.2.2 Search for reflection point
Figure out all the values of objective function f (Xj ),
and find out the worst vertex Xh which makes the value
of f (Xj ) maximum, Xc is the centre of the vertexes that
do not include the worst vertex Xh .
Define a reflection coefficient ( 1), the reflection point Xa can be defined by following formula:
Checking whether Xa is valid, if not, reducing the

value of to its half, then deal Xa with formula (7)
until it becomes valid.
2.2.3 Compare the value of f (Xj ) at Xa with that at

Xh
There are two possible results:
If f (Xa ) < f (Xh ), i.e. Xa is better than Xh , replace Xh
with Xa , the new complex is formed, then, turn to step
2 and go on.
If f (Xa ) f (Xh ), i.e. Xa is not better than Xh , reduce
the value of to its half again until Xa is better than
Xh , then turn to step 2 and go on.
2.2.4 Criterion of convergence
There are many criteria of convergence, but the criterion used most widely is that the values of f (Xj )
at all vertexes can conform to constraint condition as
follows:
where f (X s ) is the value of objective function at the

centre point; is a little positive number.
Find the vertex at which the value of objective
function is least, the optimum vertex is found.
3
3.1
OPTIMIZATION OF SECTION SHAPE AND

LINING OF ROAD TUNNEL
General situation of the road tunnel
The Bai Yang Chong No. 1 tunnel is a twin tunnel connected with arc mid-wall which is located in Jiangxi
province, China. It is 215 m long. The longitudinal
grade of this tunnel is 2.499%. The thickness of the
overburden ranges from 10 m to 20 m. The stability
of the surrounding rock is not good. The underground
water condition is well. The width of tunnel is about
11m, and the height is about 5 m.
3.2 The optimization problem of road tunnel
Lining is the most important component of tunnel
structure. The stress condition of lining is the key influence factor of the stability and the life-span of tunnel,
it usually becomes the optimization object when tunnel is designed. When optimizing the tunnel section,
the stress condition of lining should be analyzed at
first. The most important influence factor on the stress
condition of lining becomes the optimization object.
With the Tongji GeoFBA numerical analysis soft
(a type of numerical analysis soft which can solve
two-dimensional problem with finite element method,
and it is developed by the Department of Geotecnical
Engineering of Tongji University), we can get the inner
force of the tunnel lining as follows:
According to Figure 1 and quantities of calculation results, it can be found that the maximum inner
forces (including bending moment, shear force and
526
Figure 2. Flow chart of the optimization design.
the stress condition is greatly improved, and the cost

is effectively reduced.
Figure 1. Inner force diagram of tunnel lining.
3.3 General process of the optimization design

axial force) occur at the tunnel arch spring. The section shape of this tunnel is monocentric circle, and
the invert arch is joined by arc, i.e. the spring is the
linkage between the invert arch and the arch ring. It
is the most critical part of the lining ring. Due to this,
the safety factor of the lining is reduced. The reason
is that the spring is the link of two arcs with different
radius, and its radius is small, the stress concentration
occurs easily. When optimizing the section form of the
tunnel, the dimension parameter of this part is taken
as the design variable, and the maximum inner force
of lining becomes the control objective.
Tunnel lining takes about 50% of the total cost.
After improving the stress condition by optimizing the
section shape, the thickness and reinforcement of the
tunnel lining can then be optimized to reduce the lining
cost.
Generally, with an engineering example, the optimization of section shape and lining design of road
tunnel is studied in this paper. By optimization design,
There are two steps of the optimization design. First,

optimize the section shape of the tunnel, then, optimize
thickness and reinforcement of the tunnel lining.
3.4 Optimization of the tunnel section shape
3.4.1 The building of optimization model
3.4.1.1 Selection of the design variable
As shown in Figure 3, to define the tunnel section, there
are 4 independent variables: R1 , R2 , 1 , 2 . According
to the geometrical relationship, R3 , 3 can be defined
as follows:
The stress concentration occurs at the linkage part,

the dimension parameters of this part R2 , 2 is taken
as the design variables, R1 , 1 is taken as constant.
527
where R1 , 1 are constants; R2 , 2 are design variables;

R3 , 3 can be defined by formulas (9) and (10).
The upper limit and the lower limit can be defined
according to the practical situation.
3.4.2 Implement of the optimization process
Complex method is used as the numerical optimization method, programmed with C++ language, and
the optimum solution of the design variables can be
gained by running the C++ program.
3.4.3 Optimization design and comparative
analysis
Define the values of every design variable according
to the optimum solution.
The initial design:
The optimization design:
Figure 3. Sketch of tunnel section.
where R1 is the radius of roof; R2 is the radius of the

linkage part; R3 is the radius of invert arch; 1 is the
angle between R1 and springing line; 2 is the central
angle of the linkage part; 3 is the angle between R3
and the center line of tunnel section.
3.4.1.2 The building of the objective function
The radius of the linkage part is small, the curvature is
big, this is the key factor of the stress concentration.
So, the curvature of the linkage part is taken as the
objective function, the optimization problem is getting
its minimum value.
Objective function is shown as follows:
3.4.1.3 The building of the constraint conditions

R1 , 1 are constants, the inner contour line of the tunnel
section will not enter into the construction boundary. In
order not to increase the excavation cost, the minimum
excavation area should not be larger than that of the
initial design.
The minimum excavation area can be defined as
follows:
The optimization of road tunnel is realized by the

Tongji GeoFBA numerical analysis soft, the inner
forces of the tunnel lining can be gained after the
optimization design as follows:
As shown in Figure 4, the maximum inner force also
occurs at spring, but its value has been reduced a lot.
The stress condition has been improved greatly by the
optimization design.
From Table 1, we can find that the stress condition of tunnel lining has been improved, especially the
bending moment which is the determining factor of the
lining safety has been reduced greatly. It illustrated that
decreasing curvature of the linkage part is an effective
measure to improve the stress condition.
3.5 Optimization of the tunnel lining
3.5.1 The building of optimization model
3.5.1.1 Defining of objective function and design
variable
After the optimization design of the tunnel section
shape, the stress condition has been improved greatly,
then, we can optimize the tunnel lining, and take the
lining cost as the objective function.
Cost of the steel concrete lining is composed of the
cost of steel and the concrete. The cost of lining can
be defined as follows:
Then the constraint condition can be gained as follows:

where Cc is the cost of concrete; Cs is the cost of steel;
Z is the objective function.
528
3.5.1.2 Building of the constraint conditions

1) Constraint condition of shear resistance
Check the safety of the section where the shear force
reaches the maximum value.
where K is the safety factor, it can be defined according to the standard code; Q is the maximum shear
force of the tunnel lining; Ra is compressive strength
of concrete; b is the section width, which is equal to
1000 mm; ho is the design variable.
2) Constraint condition of compressive and bending
resistance
Check the safety of the sections where the axial
force reaches the maximum value, or the bending
moment reaches the maximum value. Then we can get
the following equations.
In big eccentric compression condition:
In small eccentric compression condition:
or
3) Constraint condition of boundary

The upper limit and the lower limit of the lining
thickness are determined according to the criterion.
The upper limit and the lower limit of the reinforcement area are determined according to the maximum
reinforcement ratio and the minimum reinforcement
ratio.
Figure 4. Inner force diagram of tunnel lining of the

optimized section shape.
Table 1. Inner force of tunnel lining comparing between the

optimization design and the initial design.
Initial design
Optimization design
Reduction ratio
Mmax
(kN.m)
Nmax
(kN)
Qmax
(kN)
311.1100
191.6200
38.41%
715.7400
578.9500
19.11%
285.8700
221.4300
22.54%
3.5.2 Implement of the optimization process

Complex Method is used as the numerical optimization method, C++ language is used to program and
gain the optimum solution of the design variables.
Multiplying the quantity of consumption by the unit

price can get the cost of the materials. As the section
shape is already determined, Cc is only related to h
(thickness of the tunnel lining), as the thickness of
the protective concrete layer is a constant, Cc is only
related to ho (effective section height). The longitudinal reinforcement is arranged as the constructional
reinforcement, the hoop reinforcement is the main
working reinforcement. So ho and Ag (area of the hoop
reinforcement of lining per meter) are taken as the
design variables of the objective function.
3.5.3 Optimization design and comparative

analysis
The values of every design variable can be determined according to the optimum solution, then the
comparison can be done between initial and optimized
design.
Initial design: h = 550 cm, Ag = 1571 mm2
Optimized design: h = 450 cm, Ag = 1451 mm2
where h is the thickness of the tunnel lining; for
symmetric arrangement, Ag is the hoop reinforcement
area of one side of the tunnel lining.
For this tunnel, the lining cost of the optimized
design can be reduced by 12.1%, and 630,000RMB
can be saved. It can conclude that optimization design
is an effective measure to reduce the engineering
cost, while the safety of the structure can be ensured
simultaneously.
529
ACKNOWLEDGEMENTS
CONCLUSION
From optimization design of the road tunnel and the

comparative analysis, conclusions can be gotten as
follows:
1. When defining the objective function and the
design variables, many factors can be taken into
account, one step optimizing not only makes the
problem complex, but also get no effect results.
By dividing the optimization problem into several
steps, the optimization model and operation process can be simplified greatly when solved step by
step, it can get good and practical results.
2. By numerical computation and analysis, it can be
found that the maximum inner forces occur at the
tunnel spring, the reason is that the curvature of this
part is small, and the stress concentration occurs
easily, reducing the curvature of this part is an effective measure to improve the lining stress condition,
and optimized section shape can then be gained.
3. Introducing optimization method into structure
design can reduce the construction cost effectively
without lowering the safty grade, it should be used
widespread.
4. To solve the non-linear optimization problem with
inequation constraint condition, Complex Method
is an effective optimization method, its theory is
simple, and the operation is easy and effective.
This paper are sponsored by The National High Technology Research and Development Program (863
Program) of China, No. 2006AA11Z118, Popularization Project of Key Research Technology of Twin
Tunnel Construction and Shanghai Leading Academic
Discipline Project, Project Number: B308.
REFERENCES
Ding, W.Q., Yue, Z.Q., Tham, L.G., Zhu, H.H., Lee C.F. &
Hashimoto, T. 2004. Analysis of Shield Tunnel. International Journal for Numerical & Analytical Methods in
Geomechanics. 28: 5791.
JTG D70-2004. Road tunnel design code.
Liu, Y.J. & Gao, G.F. & Feng, W.X. 2004. Study on optimization design of cross section of the large-span highway
tunnel. Liao Ning communication science and techlonogy.
(2) : 4850.
Qian, N. 1999. C++ program design course. Bei Jing: Qing
Hua University Press.
Zhang, B.H. 1998. Civil structure optimization design. Shang
Hai: Tongji University Press.
530
Floor heave behavior and control of roadway intersection in deep mine

B.H. Guo & T.K. Lu
School of Energy Science & Engineering, Henan Polytechnic University, Jiaozuo, Henan, P.R.China
ABSTRACT: For investigating floor heave behavior and controlling technique of roadway intersection in deep
mine, creep deformation characteristics of floor around roadway intersection was studied by Flac3D and the
effects of shear plastic critical value, reinforcement measurement on creep deformation of floor around roadway
intersection were discussed. As a result, the creep deformation curve can be divided into two stages including
initial creep deformation stage and softening deformation stage, which can explain the case that deformation of
roadway intersection is little before a certain time, but then increases at a high speed; only when total shear plastic
value exceeds the critical value can second creep stage take place; reinforcement measurement with bolting in
roof and ribs has little effect on floor heave, yet exerting pressure against floor can reduce floor heave obviously.
INTRODUCTION
Rocks mass of roadway in deep mine show soft behavior because they are located in high stress environment.
Due to its larger cross section and complex geometries
deformation of floor around roadway intersection is
usually greater than roadway doesnt intersect. Floor
heave has a great influence on ventilation, transportation, and so on, thus a lot of literatures already exist
about this issue, but there are only a few in which
was based on alternating effects between softening and
creeping (Yang et al. 2006).
Strata of North ventilation roadway at 990 m level
in Tangkou colliery is silty mud rock in mainly green
and cinereous, mingled with a little fine sand rock at
some locality. Protodrakonov scale of hardness f is
2.2, pressive strength is 19 MPa, and tensile strength
ranges from 1.67 MPa to 4.45 MPa. The burial depth
of roadway is nearly from 1029 m to 1035 m, maximum main stress is about 31.6 MPa, and the vertical
stress is between 25 MPa and 26 MPa ( Liu et al. 2005),
which is larger than its pressive strength, thus softening deformation and creeping deformation will occur
simultaneously (Wang et al. 1994). A case has been
observed that the deformation of roadway intersection was less before about the 45th day after it was
formed, but then accelerated and convergence between
roof and floor reached 565 mm at the 90th day, among
which floor heave accounted for more than 70 percent. Therefore, behavior and controlling technique of
roadway intersection in deep mine was investigated
according to conditions of a roadway intersection at
990 m level in Tangkou colliery, with effects of
softening/creeping of rocks mass interaction being
considered by means of numerical simulation method.
NUMERICAL SIMULATIONS PROCEDURE
2.1 Failure criteria

Pwipp model (a visco-plastic model combining WIPP
model (the rock creep visco-elastic model) and the
Drucker-Prager model) is used in this study. Total
strain includes deviate strain and average strain. Deviated strain contains three components of elastic strain,
plastic strain and viscosity strain, while average strain
contains two components mentioned above except for
viscosity strain (Liu et al. 2005).
Strength of rocks mass decreases gradually along
with development of deformation after rocks mass
begin to failure (Wang et al. 2006), and the bearing
capacity of rocks mass in plastic regions is lower than
in elastic regions. Cohesion and friction angle lower at
different degrees (Xiao et al. 2005, Zhang et al. 2005)
along with development of plastic deformation, residual cohesion will lost totally and bearing capacity is
provided by only friction force (You 2005). Therefore,
strain-softening model can be used to study deformation of roadway intersection undoubtedly (Yang et al.
2002).
Change of strain especially plastic strain externalizes the loading path and history, and reflects softening
process of material from initial condition to final failure (Zheng 2007). Referring to relationship between
softening process and plastic deformation (see Fig.1)
from You (2000), Diagrammatic sketch of full shear
plastic strain-softening curve (see Fig. 2a) was got.
Figure 2a shows that the softening coefficients of
strength k (the ratio of stress corresponding to plastic strain to strength peak value) increases with shear
plastic strain when shear plastic strain is below p1 ,
531
40
30
confining
pressure
30MPa
150
1(MPa)
(13) MPa
200
20
100
10
50
10
10
0
0
0.4
0.8
1.2
p/mm
1.6
k1
10
1(103)
Figure 1. Relationship between softening process and plastic deformation (after You, 2000).
20
Figure 3. Stress-strain curves under cycle loading (after

Zhu, 1985).
k0
0
p1
p2
(a)Full curve
pc
(b)Simplified curve
Figure 2. Diagrammatic sketch of softening coefficientshear plastic strain curve.
decrease when the shear plastic strain is between p1

and p2 , and keep unchanged as a constant value k0
when shear plastic strain is above p2 . Because p1 and
k0 are usually very little, they are both assumed as 0
here, thus the simplified shear plastic strain-softening
curve (see Fig. 2b) was obtained. Besides, when
total strain of rocks mass exceeds the strain value at
peak strength, deformation modulus lessen gradually
with strain, but residual deformation modulus is never
reach 0 (see Fig. 3).Other researchers (Zeng et al. 2005,
Liang et al. 2005, Cheng et al. 2005, Chen et al. 2005,
Li et al. 2006, Qiang et al. 2006, Wang
et al. 2007) also reported the similar opinion.
If we define k1 as strength softening coefficient and
k2 as deformation softening coefficient, they can be
defined by following formulas:
Figure 4. Simulated model.
where p = shear plastic value; and pc = shear plastic

critical value, 2e-3 here.
2.2
Model development
Assumption of rocks mass being homogeneous and

in hydrostatic-pressure state was made in this work.
Simulated model containing a T shaped roadway intersection and bolt supporting sketch were plotted in
Figure 4 and Figure 5 respectively. The sizes of calculated model were 40 m length, 40 m width and 40 m
height. The bottom of the model was fixed in all directions, four sides were fixed in horizontal direction, and
overburden weight was exerted on the top of the model.
The section of roadway was rectangle with width of
4 m and height of 3 m. The length and diameter of 2
anchors bolted in the roof were 6300 mm and 17.8 mm
respectively, the interval along the axial and circumferential direction of the roadway were 3000 mm and
2400 mm respectively; for cables, the corresponding
values were 2300 mm, 18 mm, 1000 mm and 1200 mm
532
Table 3.
Figure 5. Bolt supporting sketch.

Table 1.
Mechanical parameters of cables and anchors.
Parameters
Elastic
modulus
MPa
Grout
cohesion
strength
MPa
Grout
exposed
perimeter
m
Cable
Anchor
45
195
0.2
0.42
1
1
Parameters
Tensile
strength
MPa
Grout
stiffness
MPa
Cross-section
area
m2 106
Cable
Anchor
0.25
1.85e3
17.5
5.35e3
254
249
Mechanical parameters of rocks mass.
Bulk
Shear
Shear
Tensile
modulus modulus Density strength strength
Parameters K
G
D
f
t
3
Units
Values
Table 2.
GPa
8
GPa
4.8
kg/m3
2640
MPa
9.5
J mol1
50160
WIPPModel
constant, D
J mol1 K1 Pan s1
1.987
28.95e-36
K
300
Parameters
WIPP-model
constant, A
WIPP-model
constant, B
Values
22.8
WIPP-model
exponent, n
25.4
4.9
Parameters
Material
parameter,
q
Material
parameter,
qk
Critical
steady-state
creep rate
Values
0.55
0.5
1.078 e-8
RESULTS
3.1 Analysis of creep stages
Creep parameters of rocks mass.
Activation Zone
Gas
energy,
temperature, constant,
T
R
Parameters Q
Units
Values
MPa
3.06
respectively. The cables near corner deviated 20 from

normal direction of wall to corner. According to relative references (Liu, J.H et al. 2005, Liu, T.S et al.
2005), mechanical parameters and creep parameters
of rocks mass were chosen and listed in table 1 and
table 2; mechanical parameters of cables and anchors
were listed in table 3. Following 4 reinforcement measurement methods were investigated by numerical
simulation: method A was for naked roadway intersection, method B and C were for bolted roadway
intersection without and with cables applied in floor at
two corners, and the last method was method C companied with exerted pressure against floor, the pressure
values ranged from 0.1 MPa to 1.0 MPa with interval
of 0.1 MPa.
XU et al. (2007) discussed that deformation process of the roadway can be divided into three stages
including adjustment deformation stage, stable defor
mation stage and accelerated deformation stage by
investigation in the field.
Figure 6 shows deformations of naked roadway
intersection versus creep time curves under different
critical values. When pt > pc , Figure 6a illustrates
that creep curves can be divided into two creep stages
named initial creep stage and softening creep stage
respectively, and each stage include a decelerated
creep stage and a stable creep stage. At initial creep
stage where the shear plastic value is lower than
the critical value, deformation increases rapidly and
then comes to a nearly constant value until shear
plastic value exceeds the critical value. At softening
creep stage, deformation of the roadway intersection
increases rapidly with creep time for a longer period,
and comes to stable state finally. Therefore, we can
explain the case mentioned in section 1 that the deformation is little before a certain time, but increases at a
high speed later.
For naked roadway intersection, when pt < pc ,
only initial creep stage occurs (see Fig. 6b). Roof
deformation raises rapidly with a gradual decreasing
velocity, then approaches a relatively constant value
while the deformation velocity come to about 0; floor
heave raises rapidly also, but it reaches a peak value,
then decreases a little, finally come to a lower constant value. There is a deformation rebound in the
deformation process of floor heave.
When pt > pc , and roadway intersection is reinforced by bolting and exerting pressure against floor,
the relationships between deformation and creep time
are shown in Figure 6c. Figure 6c illustrates that the
533
Figure 7. The relationships between deformations, vertical

stress peak value and reinforcement measurement: A, B and
C represent Method A, method B and method C described in
section 2.2; 0.1 1.0 MPa refer to the pressure values exerted
against floor based on method C.
3.2 Controlling technique
Figure 6. Deformation-creep time curves under three conditions, pt represents total shear plastic train value.
floor heave is very little compared with roof deformation; the profile of roof deformation is similar with that
shown in Figure 6a, yet floor heave curve is very different. The beginning time of softening creep stage takes
place late compared with that of roof deformation,
and its total deformation value is less greatly than that
of roadway intersection without reinforcement shown
in Figure 6a. Therefore, Reinforcement for floor can
reduce floor heave effectively.
Figure 7 illustrates that roof deformation decreases

abruptly when naked roadway intersection are reinforced by bolting in roof and ribs, yet other reinforcement measurements have little affection on it; floor
heave decreases a little when naked roadway intersection are reinforced by bolting in roof and ribs, even
when cables were applied in floor near two corners
(see Fig.7a), but floor heave lessens obviously with
increment of pressure value exerted on the floor, and
vertical stress peak value around roadway intersection also decreases (see Fig.7b); vertical stress peak
values around roadway intersection with reinforcement measurement are all lower than naked roadway
intersection. Only when the pressure value against
floor is 0.1 Mpa and with reinforcement measurement
method C, the vertical stress peak value has a little
rebound. Therefore, we can reduce floor heave amount
by exerting pressure against floor together with bolting support. If it is not enough yet, measurements of
vertical cutting in floor (Guo, unpubl.) and so on can
be used additionally.
534
Figure 9. Distribution of higher vertical stress and softening

zone distributions under three conditions.
Figure 8. Relationship between creep time, total deformation and shear plastic critical value.
DISCUSSION
4.1 Affection of shear plastic critical value

It is very important to choose reasonable shear plastic
critical value when numerical simulation is performed.
Figure 8 illustrates that when the critical value is above
2e-3, the second creep stage doesnt occur, and the
deformation of the roadway intersection is very little,
and the total creep time that calculation last is shorter.
Otherwise, the second creep stage takes place. At the
second creep stage, the deformation of the roadway
intersection is larger and the deformation value of each
critical value seems to be similar. But along with increment of the critical value, the beginning and end time of
the softening creep stage delay a little, and total deformation lower a little. In summary, the shear plastic critical value determines if the softening creep stage takes
place, and has a key influence on the beginning time,
end time and total deformation of the second creep
stage. So creep deformation of roadway intersection
can be reduced by improving the shear plastic value.
4.2
enough. Why can roadway intersection come to stable

state finally?
Distribution of shear plastic zones (i.e. softening
zone (Wang et al. 2004)) and vertical stress filed are
plotted in Figure 9. As the results shown in Figure 9,
the final shear plastic zone is small when pt < pc (see
Fig. 9a), greater when pt > pc (see Fig. 9b), and can be
reduced by reinforcement measurements (see Fig. 9c).
When rocks mass in the softening zone lost their bearing capacity to some extent, adjacent rocks mass must
provide more bearing capacity, if they cant, they will
be softened also, and overburden pressure will continue seeking another rocks mass until the rocks mass
can bear it sufficiently and keep stable finally (see
Fig. 9d). Along with softening zones generation and
enlargement, peak stress value in surrounding rocks
mass increases and its location diverts to rocks mass
that hasnt been softened. All in all, although softened
rocks mass adjacent to excavated room lost bearing
capacity at a different degree, but rocks mass adjacent
to softened rocks mass provide higher bearing capacity, so that the construction of roadway intersection do
not lost its stability completely.
5
CONCLUSIONS
The conclusions of this research are summarized

below:
Generation mechanics of creep stages
Research workers (Zhu et al. 2002, Wang et al. 2004,

Li et al. 2006, Fan 2007) considered that rocks samples will failure finally if shear plastic value is large
1. Pwipp model of flac3D software can be used effectively to simulate creep deformation of roadway
intersection in deep mine. Strain-softening used
535
in this study includes softening of strength and

deformation modulus.
2. Creep deformation-creep time curves can be
divided into two stages including initial creep stage
and softening creep stage, and only when total shear
plastic value exceeds shear plastic critical value can
softening creep stage occur. Determination of the
critical value is very important in calculation, and
creep deformation of roadway intersection can be
reduced by means of improving the shear plastic
value.
3. Bolting in surrounding rocks mass and exerting
pressure against floor can reduce deformation of
roadway intersection in deep mine. Method of
exerting pressure against floor has more obvious
effects to reduce floor heave than other measurements used in this research, if it is not enough, other
measurement can be used additionally.
ACKNOWLEDGEMENTS
This work was carried out under the Outstanding Talent Innovation Fund of Henan province (No.
0621000400).
REFERENCES
Chen, Z.J. & Yang, J.W. 2005. Measures for supporting deep
High Stress crack-expansion creep rock mass in Jinchuan
mining district. Metal Mine (1): 1822.
Cheng, R.H., Qian, M.G. & Miao, X.X. 2005. Numerical simulation in mining pressure control of thick and
strong stratum caving by water-infusion softening method.
Chinese Journal of Rock Mechanics and Engineering
24(13):22662271.
Fan, Q.Z. 2007. Experimental study on creep and its disturbed
effect of rocks. Chinese Journal of Rock Mechanics and
Engineering 26(1):216216.
Li, Y.J., Pan, Y.S. & Zhang, M.T. 2006. Time effect on zonal
disintegration process of deep rock mass. The Chinese
Journal of Geological Hazard and control 17(4): 119122.
Liang, Z.Z., Yang, T.H. & Tang, C.A. et al. 2005. Threedimensional damage soften model for failure process of
heterogeneous rocks and associated numerical simulation.
Chinese Journal of Geotechnical Engineering 27(12):
14471452.
Liu, J.H., Zhu, W.S. & Li, S.C. et al. 2005. Analysis of rheological characteristics and stability of surrounding rock
masses of Xiaolangdi hydrojunction underground caverns
by using Flac3D. Chinese Journal of Rock Mechanics and
Engineering 24(14): 24842489.
Liu, T.S., Zhao, J.Z. &Yang, J.E. et al. 2005. Support technology of bolt and anchor truss applied in large chamber of
Tangkou mine. Coal Science andTechnology 33(5):1215.
Qiang, H., Zhou, H.Q. & Chang, Q.L. 2006. Regress analysis
of mechanics characteristic of the rock strength at the prepeak and post-peak. JiangXi Coal Science & Technology
(2): 4850.
Wang, H.P., Gao, Y.F. & Li, S.C. 2007. Uniaxial experiment
study on mechanical properties of reinforced broken rocks
pre-and-post grouting. Chinese Journal of Underground
Space and Engineering 3(1): 2731, 39.
Wang, J.A., Jiao, S.H. & Xie, G.X. 2006. Study on
influence of mining rate on stress environment in surrounding rock of mechanized top caving mining face. Chinese Journal of Rock Mechanics and Engineering 25(6):
11181126.
Wang, J.J., Lu, Z.Y. & Liu, X.F. 2005. Discussion on floor
upheaval mechanism of mine soft rock roadway. Coal
Engineering (9): 6768.
Wang, L.G., Wang Y.J. & Liu, X. 1994. The rheological instability theory for rock sample and its criteria. Journal of
Fuxin Mining Institute (Natural Science) 13(3):9397.
Wang, X.Q., Yang, L.D. & Gao, W.H. 2004. Creep damage
mechanism and back analysis of optimum support time
for soften rock mass. Chinese Journal of Rock Mechanics
and Engineering 23(5): 793796.
Xiao, H.F., He, X.Q. & Feng, T. et al. 2005. Research on
coupling laws between EME and stress fields during
deformation and fracture of mine tunnel excavation by
Flac3D simulation. Chinese Journal of Rock Mechanics
and Engineering 24(5):812817.
Xu, X.L., Zhang, N. & Xu, J.G. et al. 2007. Principle and practice of process control over soft broken roadway with high
ground stress. Journal of Mining & Safety Engineering 24
(1):5155.
Yang, C., Chui, X.M. & Xu, S.P. 2002. Establishment and
study of strain-softening numerical constitutive model for
soft rock. Rock and Soil Mechanics 23 (6):695697.
Yang, C.H. & Li, J.G. 2006. Analysis of creeping mechanisms
of non-uniformity soft rocks. Journal of Mining and Safety
You, M.Q. 2005. Study of deformation and failure of rock
based on properties of cohesion and friction. Journal of
Geomechanics 11(3):287291.
You, M.Q. 2000. Strength of rock specimens and process of
deformation and failure. Beijing: Geology Press.
Zeng, S., Yang, S.J. & Man, C. et al. 2005. Statistical
constitutive model for limestone rock damage under uniaxial compression and its experimental study. Journal
of Nanhua University (Science and Technology) 19(1):
6972, 95.
Zhang, F., Dong, Z.H. & Ding, X.L. 2005. Numerical analysis of excavation process for tailrace tunnels of Pengshui
project. Journal of Yangtze River Scientific Research
Institute 22(6):5962.
Zheng, Y.R. 2007. Discussion on yield and failure of geomaterials and stability analysis methods of slope/landslide
communion and discussion summary of special topic
forum on geologic disasters in the three gorges project
region. Chinese Journal of Rock Mechanics and Engineering 26(4):649661.
Zhu, H.H. & Ye, B. 2002. Experimental study on mechanical
properties of rock creep in saturation. Chinese Journal of
Rock Mechanics and Engineering 21(12):17911796.
Zhu, Z. F. 1985. Stiff test machine. Beijing: coal industry
press.
536
Squeezing potential of tunnels in clays and clayshales from normalized

undrained shear strength, unconfined compressive strength and
seismic velocity
M. Gutierrez
Colorado School of Mines, Golden, CO, USA
C.C. Xia
Key Laboratory of Geotechnical and Underground Engineering of Ministry of Education, Tongji University
ABSTRACT: The tunnel squeezing phenomenon was first described byTerzaghi (1946) who associated squeezing mainly with clay-rich rocks. Consistent with Terzaghis original description, the focus of this paper is on
tunnels in clays and clay-rich rocks such as clayshales. The paper develops a simple procedure to predict tunnel
squeezing potential using normalized undrained shear strength, unconfined compressive strength and P-wave
velocity. A collection of a large amount of undrained triaxial test data is used to show that the undrained shear
strength of clay shales can be normalized with respect to the effective vertical stress. The normalized undrained
shear strength can then be used to predict the squeezing potential since it can be directly related to Pecks (1969)
stability factor N . Values of N provide estimates of the degree of squeezing potential, with values of N > 1
indicating potential for squeezing. The normalized undrained shear strength of clay and shales is shown to be
related to the apparent overconsolidation ratio OCR, which accounts for both mechanical and other diagenetic
pre-consolidation, following the SHANSEP procedure for clays (Ladd and Foott, 1977). To facilitate the use
of the proposed normalized undrained shear strength vs. OCR relationship, empirical equations are established
to predict the apparent pre-consolidation stress from unconfined compressive strength c and seismic P-wave
velocity. Together with the in situ effective vertical stress prior to excavation at the tunnel location, the apparent
pre-consolidation from c or P-wave velocity can be used to estimate the apparent OCR and thence the squeezing
potential. The proposed approach is compared with available field data and to existing methods to predict tunnel
squeezing potential.
INTRODUCTION
Barla (2001) defined squeezing as the large timedependent convergence during tunnel excavation. It
takes place when a particular combination of induced
stresses and material properties pushes some zones
around the tunnel beyond the limiting shear stress at
which creep starts. Deformation may terminate during construction or continue over a long period of
time. Several other definitions of tunnel squeezing
have been proposed including those by Gioda (1982),
ORourke (1984), Kovari (1988), Singh (1988), and
Aydan et al. (1993). A full review of the different
definitions is given in Barla (2001). In general, squeezing cannot always be distinguished from swelling
conditions (Steiner, 1993).
The tunnel squeezing phenomenon was first
described by Terzaghi (1946) who associated
squeezing mainly with clay-rich rocks. Consistent with

Terzaghis original description, the focus of this paper
is on clay-rich rocks such as clayshales, although
squeezing can also occur in other rock types. Several procedures have been developed to predict the
squeezing potential of tunnels. One of the first stability criteria to predict squeezing was developed by
Peck (1969) for tunnels in clays based on Broms and
Bennermarks (1967) stability criterion for open excavations. He proposed a stability number N which is
expressed as the ratio between the total vertical stress
v at the tunnel location and the undrained shear of the
clay Su :
Values of N have been correlated with observations

of tunnel stability response, and these correlations are
537
Table 1.
1969).
Stability criteria for tunnels in cohesive soils (Peck
Problems encountered
15
56
67
Tunneling without unusual difficulties

Clay may squeeze rapidly into shield void
Shear failure ahead of tunnel causes ground
movements into the face even in shield tunneling
General shear failures and ground movements
around tunnel heading cause shield contact to
become difficult; shield tends to dive
>7
summarized in Table 1. Cases where N > 5 are considered to have high degrees of rapid squeezing, and
higher values of N indicate potential for general shear
failure and large tunnel displacements close to the
tunnel heading.
For tunnels in rocks, several tunnel squeezing criteria, which are mostly empirical, have also been
proposed. Singh et al. (1992) developed a criterion
based on the Q-system of rock mass classification
(Barton et al. 1974) and the overburden height H
(in m) which separates the squeezing cases from the
non-squeezing cases:
Tunnels with overburden height greater than that

given in Eq. (2) will experience squeezing.
Semi-empirical approaches have also been proposed by Goel et al. (1995, 2000), Jethwa et al.
(1984), Aydan et al. (1993), and Hoek and Marinos
(2000). The last three criteria use a stability number,
which is a reciprocal of Pecks stability. Almost all
these squeezing criteria summarized are empirically
or semi-empirically based on direct observations of
tunnel response. In the semi-empirical approaches of
Jethwa et al., Aydan et al., and Hoek and Marinos, the
ratio of capacity and load (or strength and stress) is
analogous to the factor of safety FS (or the reciprocal if the load to capacity ratio is used). In addition to
linking the degree of squeezing to the factor of safety,
the advantage of using stress and strength is that these
can be related to strain levels, provided the full stressstrain curve are known, which in turn can be related to
degree of squeezing in the tunnel.
The main challenge in the use of these semiempirical approaches is in the determination of the
rock mass strength. In Pecks approach, the strength is
expressed in terms of the undrained shear strength,
while in Jethwa et al., Aydan et al., and Hoek and
Marinos, the strength is expressed in terms of the
unconfined compressive strength of either the intact
rock or the rock mass. It should also be noted that the
proposed methodologies have been developed mainly
for clays or hard rocks. Fewer studies have investigated

the applicability of the above mentioned criteria to
intermediate materials such as hard soils or soft rocks
(e.g. Hoek and Marinos, 2000). Based on these observations, the main objective of this paper is to propose
simple methods to estimate the shear strength of intermediate soil-rock materials, particularly cemented
clays and shales, and to use the simple methods to estimate squeezing potential in tunnels. Estimates of the
shear strength of shales are obtained from a database of
triaxial test data on several clayshales. These estimates
of shear strengths are then linked to field observations
of squeezing to develop simple procedures for the preliminary investigation of squeezing in tunnels in clays
and shales.
2
NORMALIZED UNDRAINED SHEAR

BEHAVIOR OF CLAY AND SHALES
A procedure that is widely used to characterize the

undrained shear strength of clays is the SHANSEP
(Stress History and Normalized Soil Engineering
Properties) procedure developed by Ladd and Foott
(1977). According to this procedure, the undrained
shear strength Su of normally consolidated (NC) clays
normalized with respect to the current effective vertical stress v is unique, and for overconsolidated (OC)
clays, the following relationship adequately represents
the normalized undrained shear strength:
where OCR is the overconsolidation ratio defined as

the ratio between the maximum past effective vertical
stress p and the current effective vertical stress v ,
that is:
The parameter b is an empirical exponent, and

a = (Su /v )NC is the normalized undrained shear
strength of NC clay, i.e., the value of Su /v for
OCR = 1.
Numerous studies in the literature have shown
the applicability of SHANSEP in representing the
undrained shear strength of many types of clays,
including marine, residual and glacial soils. Although
some studies have shown the successful use of
SHANSEP for clays with some degree of cementation
(e.g. Bo et al. 2003), the applicability of SHANSEP
to lithified materials like shales has not been fully
established.
SHANSEP has been successfully used in practice
in geotechnical analysis and design, and it would
538
Figure 1. Normalized undrained shear strength as function

of apparent OCR for 25 different types of shales.
Figure 2. Correlation between apparent preconsolidation

and unconfined compressive strength for different shales.
be valuable to extend the procedure to shales and

highly cemented cohesive materials. The main difference with clays is that overconsolidation and shear
strength in shales are effected not only by mechanical loading-unloading, but also by the other diagenetic
processes particularly cementation at the clay particle contacts. The increase in overconsolidation due to
non-mechanical processes is called apparent or quasi
preconsolidation (e.g. Bjerrum and Wu, 1960).
To investigate the applicability of SHANSEP to
shales, a database of triaxial test results on 25 types
of clayshales from different locations was assembled.
Included in this database are consolidated undrained
(CU) triaxial test results on shales with different consolidation stresses. The normalized undrained shear
strength of 25 types of materials are plotted against
apparent OCR in Fig. 1. Only clayshales, which are
shales containing more than 50% clay particles by
weight, and stiff cemented clays are included in the
study. The apparent OCR is defined as ratio of the
current effective vertical stress and the apparent preconsolidation stress, i.e. OCR = y /v , where y is
simply the effective vertical stress at which yielding
can be observed from the experimental consolidation
stress-strain curve. In case of uncemented materials,
y = p .
The results show a linear relationship between the
log(Su /v ) and log(OCR), which agrees with the power
function given in Eq. (3). The reasonably good correlation between Su /v and OCR for 25 types of materials
is very promising, and indicates that SHANSEP can
provide a reliable approach to predicting the undrained
shear strength of shales. The average values of a and b
for the 25 different materials are equal to 0.37 and 0.87,
respectively. In comparison, Ladd and Foott (1977)
obtained values of a = 0.20 and b = 0.77 for several

clays from CU triaxial tests. It is important to note that
the data shown in Fig. 1 are for clayshales with different degrees of diagenesis and cementation. Although
the normalized undrained shear strength of shale samples are similar, more lithified samples are actually
stronger (i.e., have higher undrained shear strength Su )
than younger uncemented samples because of higher
values of the apparent preconsolidation stress.
3
DETERMINATION OF THE APPARENT

PRECONSOLIDATION STRESS IN SHALES
The most widely used laboratory approach to determine y experimentally is Casagrandes (1936) procedure where y corresponds to the sharpest bend in
consolidation plot. In addition to this approach, it is
useful to develop procedures to estimate the apparent preconsolidation stress from simple index tests to
facilitate the application of SHANSEP to shales. One
parameter that can be obtained with relative ease in
the field or in the laboratory is the unconfined compressive strength c . In the following, it is assumed
that the intact rock and rock mass c are the same, as
was done by Aydan et al. (1993), and Hoek and Marinos (2000). Figure 2 presents a plot of y vs. c (both
in MPa) for different clayshales, which shows a reasonable correlation between the two parameters of the
following form:
Although there is some scatter in the data, they

are in the range of typical c values of about 1 to
539
SQUEEZING POTENTIAL OF TUNNELS IN

CLAYS AND SHALES
Substituting Eq. (3) in Eq. (1) yields the following

expression for the stability number:
The total and effective vertical stresses are related

to the depth to the tunnel H and the total unit weight
and the effective unit weight of the material above
the tunnel, i.e. v = H and v = H . Both these unit
weights depend on the specific gravity of the solids
Gs , the porosity n of the material, and the degree of
saturation. Assuming fully saturated conditions, the
ratio v /v can be expressed as:
Figure 3. Correlation between apparent preconsolidation

and compressional wave velocity for different shales.
70 MPa for shales. It should be reasonable to expect

that a correlation exists between y and c since
both can be considered as material parameters, which
depend on the degree of mechanical consolidation and
cementation.
A more convenient procedure is to relate c with
data that can be directly measured in the field using in
situ tests such as the compressional wave velocity Vp .
One such correlation obtained by curve-fitting through
experimental data, shown in Fig. 3 in terms of y (in
MPa) and compressional wave velocity Vp (in m/s),
also assumed the same for the intact rock and the rock
mass, is obtained as
Re-writing Eq. (8)
and solving for the effective vertical stress gives a relationship between the effective vertical stress and the
apparent yield stress:
In turn, substituting Eqs. (5) and (6) in the above

equation results in the following relationships between
the effective vertical stress, and the unconfined compressive strength and the P-wave velocity:
Again, data are limited but are within the typical
range of Vp of 2 to 5 km/s for shales. To support
the reliability of Eq. (6), an additional relationship
is formulated via the following correlation for shales
developed by Urai (1995) between c (in MPa) and Vp
(in m/s):
Substituting Eq. (7) in (5) provides the second y vs.

c curve shown in Fig. 3, which is in close agreement
with Eq. (6). The similarity of the two curves shown
in Fig. 3 appears to support the correlation between y
and Vp . Obviously more data are needed, but Eqs. (5) to
(7) can provide preliminary estimates of the apparent
preconsolidation stress in the field.
Equations (12) and (13) may be viewed as stability criteria which relate the effective vertical stress
corresponding to the tunnel depth v , the unconfined
compressive strength c or P-wave velocity Vp , the
stability number N , the porosity function F(n), and
the empirical constants a and b. Tunnels with v ,
larger than those given in Eqs. (12) and (13) have high
potential for squeezing.
540
Figure 4. Comparison of the stability criterion given in

Eq. (14) with observed cases of squeezing and non-squeezing
in tunnels in clays (data from Broms and Bennermark, 1967).

in tunnels in clayshales (data from Bhasin, 1991).
a = 0.37 and b = 0.87 taken from Fig. 1. Substituting

these values in Eq. (12) gives the following equation:
COMPARISON WITH FIELD DATA AND

OTHER EMPIRICAL TUNNEL SQUEEZING
CRITERIA
To show their validity, the stability criteria given in Eqs.

(12) and (13) are compared with field data on response
of tunnels in clays and clayshales. The validity of Eq.
(12) for tunnels in clays is shown in Fig. 4, which
shows data taken from Broms and Bennermark (1967).
The data have been re-plotted in terms of the effective
vertical stress and unconfined compressive strength.
A boundary curve between the non-squeezing and
squeezing cases is established using Eq. (12) in conjunction with a value of F(n) 2.5 (based on typical
values of n 60% and Gs = 2.7 for clays), and a stability of number of N = 5.3. Also, values of a = 0.20 and
b = 0.77 were used, which are representative values
for clays given by Ladd and Foott (1977). Substituting these values in Eq. (12) gives the following
criterion which demarcates cases of squeezing and
non-squeezing for tunnels in clays:
It can be seen that Eq. (14) provides a clear boundary between cases of squeezing and non-squeezing in
tunnels in clays.
For clayshales, the validity of Eq. (12) is validated
by comparison with data on Himalayan tunnels collected by Bhasin (1991). The data of Bhasin (1991)
have been re-plotted as for the data for clay and are
shown in Fig. 5. A boundary curve between the nonsqueezing and squeezing cases is established using
Eq. (12) with values of F(n) 1.8 (for typical values of n 30% and Gs = 2.7 for shales), N = 2.5, and
It can be seen that Eq. (15) provides a clear boundary between cases of squeezing and non-squeezing in
tunnels in shales.
Figure 6 shows the combined data from clays and
clayshales. A boundary curve demarcating the case
of squeezing from non-squeezing for tunnels in both
clays and shales is established from the average of the
stability criteria given in Eqs. (14) and (15). This curve
is shown in Fig. (6), which shows that Eq. (15) can
adequately separate the case histories of squeezing and
non-squeezing for tunnel in both clays and clayshales.
Figure 6 also shows the stability criterion of Singh

et al. (1992) combining Eq. (2) with the empirical relationship between c and Q-value also from Singh et al.
(1992). It can be seen from Fig. 6 that the Singhs
criterion significantly underestimates the boundary
between squeezing and no-squeezing for both clays
and shales.
The final comparison with field data is shown in
Fig. 7 to show the validity of the criterion given in
Eq. (13). The unconfined compressive strength data
of Bhasin (1991) were converted to Vp using Eq. (7).
Using similar parameters used for Eq. (15), the following stability criterion is obtained in terms of v
and Vp :
541

Eqs. (16) and (18) with observed cases of squeezing and
non-squeezing in tunnels in clays and shales (data from
Broms and Bennermark, 1967; and Bhasin, 1991).
on normalized undrained shear strength as embodied in the SHANSEP procedure for clays. It was
shown that the undrained shear strength of clayshales
also exhibit normalized behavior similar to unlithified
clays. In addition to relating stability to overconsolidation ratio, squeezing criteria in terms of unconfined compressive strength and P-wave velocity were
proposed.
The validity of the proposed criteria was demonstrated by comparison with field observed squeezing
in tunnels in clays and shales. It was shown that the
proposed criteria are capable of delineating cases of
squeezing and non-squeezing in tunnels in clays and
clay shales. The proposed criteria provide better predictions of squeezing in clays and clayshales than
other empirical criteria for hard rocks such as the one
proposed by Singh et al. (1992).
The proposed criteria can be used for preliminary
analysis of squeezing in tunnels in clays and clayshales
where there are limited data. For tunnels with sufficient
shear strength data, the paper has demonstrated the
possibility of using SHANSEP procedure in conjunction with project specific soil parameters to estimate
tunnel stability.
ACKNOWLEDGEMENT
This paper are based upon works supported by
the National Science Foundation under Grant No.
0324889 and Shanghai Leading Academic Discipline
Project, Project Number: B308. Any opinions, findings, and conclusions or recommendations expressed
in this material are those of the authors and do not
necessarily reflect the views of the National Science
Foundation.

in tunnels in shales (data from Bhasin, 1991).
As can be seen, Eq. (17) provides an adequate criterion for separating case histories of squeezing and
non-squeezing in tunnels in clay and shales. Also
shown is the criterion of Singh et al. (1992) re-plotted
in terms of Vp by using the relationship between Vp and
Q-value developed by Barton (2002). It can be seen
that Singhs criterion expressed in terms of Vp underestimates the boundary between cases of squeezing
and non-squeezing.
CONCLUSIONS
Squeezing criteria for materials that range from clays

to clayshales were proposed. The criteria were based
REFERENCES
Aydan, O., Akagi, T. & Kawamoto, T. 1993. The squeezing
potential of rock around tunnels: theory and prediction.
Rock Mech. Rock Eng. 2: l37163.
Barla, G. 2001. Tunneling under squeezing rock conditions.
Lecture Notes, Eurosummer-School in Tunnel Mechanics,
Innsbruck.
Barton, N., Lien, R. & Lunde, I. 1974. Engineering classification of rock masses for the design of tunnel supports.
Rock Mech. 6(4): 189239.
Barton, N. 2002. Some new Q-value correlations to assist in
site characterisation and tunnel design. Intl. J. Rock Mech.
Mining Sci. 39: 185216.
Bhasin, R. 1991. Evaluation of soft rock conditions in tunnels
through the Lower Himalayan regions; A Contribution for
updating of the Q-system. MSc Thesis, University of Oslo,
1991.
Bjerrrum, L. & Wu, T.H. 1960. Fundamental shear strength
properties of the Lilla Edet clay. Gotechnique, 10(3):
101109.
542
Bo, M.W., Choa, V. & Hong, K.H. 2003. Material characterization of Singapore Clay at Changi. J. Eng. Geol.
Hydrogeol. 36: 305319.
Broms, B. & Bennermark, H. 1967. Stability of clays at
vertical openings. Swedish Geotechnical Institute Publ.
No. 16.
Casagrande, A. 1936. The determination of the preconsolidation load and its practical significance. Proc. 1st
Intl. Conf. Soil Mech. Fnd. Eng., Cambridge, Mass. 60.
Gioda, G. & Cividini, A. 1996. Numerical methods for the
analysis of tunnel perform-ance in squeezing rocks. Rock
Mech. Rock Eng. 29(4): 171193.
Goel, R.K., Jethwa, J.L. & Paithakan, A.G. 1995. Tunnelling
through the young Himalayas A case history of the
Maneri-Uttarkashi power tunnel. Engrg. Geol. 39: 3144.
Hoek, E. & Marinos, P. 2000. Predicting tunnel squeezing
problems in weak heterogeneous rock masses. Tunnels
Tunnel. Intl.: pp. 4551 (part one), 3336 (part two).
Jethwa, J.L., Singh, B. & Singh, B. 1984. Estimation of
ultimate rock pressure for tunnel linings under squeezing rock conditions a new approach. In E.T. Brown,
J.A. Hudson (eds.), Design and Performance of Underground Excavations, ISRM Symposium, Cambridge:
231238.
Kovari, K. 1998. Tunnelbau in druckhaftem Gebirge

Tunnelling in squeezing rock. Tunnel 5: 1231.
Ladd, C.C. & Foott, R. 1977. New design procedure for stability of soft clays. J. Geotech. Eng. Div., ASCE, 100(GT4):
763779.
ORourke, T.D. 1984. Guidelines for tunnel lining design.
ASCE.
ground. State of the art volume, 7th Intl. Conf. Soil Mech.
Fnd. Eng., Mexico: 225282.
Singh, B., Jethwa, J.L., Dube, A.K. & Singh, B. 1992. Correlation between observed supportpressure and rock mass
quality. Tunnel. Undergr. Space Tech. 7: 5974.
Steiner, W. 1993. Swelling rocks in tunnels: Rock characerization, effect of horizontal stress and construction
procedures. Intl. J. Rock Mech. Min. Sci. & Geomech.
Abstr. 30(4): 361380.
Terzaghi, K. 1946. Rock defects and loads in tunnel supports.
Rock tunneling with steelsupports. In R.V. Proctor &
T.L. White (eds.),The Commercial Shearing and Stamping
Co., Youngstown, Ohio: 1799.
Urai, J.L. 1995. Brittle and ductile deformation of mudrocks.
EOS Transactions, American Geophysical Union, Nov. 7,
1995, F656.
543
Framework of performance-based fire protection design method for

road tunnel
X. Han
Shanghai Institute of Disaster Prevention and Relief, Tongji University, Shanghai, P. R. China
G.Y. Ding
Shanghai Tunnel Engineering Co. Ltd, Shanghai, P. R. China
ABSTRACT: Under the background of performance-based code for fire protection design of building, and on
the basis of the state-of-the-art of fire safety engineering as well as prescriptive-based code for fire protection
design of road tunnel, this paper outlines the framework of performance-based fire protection analysis and design
method for large cross-section road tunnel. The framework provide a preliminary guidance on the application of
scientific and engineering principles to the protection of such road tunnels from the unwanted effects of fire.
INTRODUCTION
With rapid development of tunnel construction in

China, the number of road tunnels has increased
sharply to serve the perpetual growth of both freight
and transport for the populace. The consequence of
the growth in the number of road tunnels and the volume of traffic is an increasing occurrence of severe
fire incidents in tunnels. Also, as tunnel construction technology has evolved, so the feasible length
and cross-section of road tunnels have expanded. Subsequently, the control of such incidents has become
more difficult for the emergency services of large
cross-section road tunnel.
Currently, the fire protection design of most
road tunnels are in accordance with prescriptivebased codes. Prescriptive-based codes provide specific
requirements for broad classifications of tunnels that
establish acceptable or tolerable levels of risk for a
variety of health, safety, and public welfare issues.
These requirements are generally stated in terms of
fixed values, such as maximum travel distance, minimum fire resistance ratings, and minimum features
of required systems (e.g. detection, alarm, suppression, and ventilation), and not in terms of overall
tunnel performance. However, the construction of
large cross-section road tunnel would challenge more
complicated fire protection requirement and should
be provided scientific and more cost-effective fire
protection solution.
This paper established a framework of performancebased fire protection analysis and design method
for large cross-section road tunnel. The framework
provided preliminary guidance on the application of

scientific and engineering principles to the protection of such road tunnels from the unwanted effects
of fire. It also constructed a process for undertaking a
performance-based fire safety engineering approach
to road tunnel fire safety analysis and design. The
method requires the use of a variety of tools in the analysis, bringing increased engineering rigor and often
resulting in innovative design options. It allows the
safety levels provided by alternative design options
to be compared. In this way, it could result in a
comprehensive fire protection strategy for large crosssection road tunnel in which all fire safety systems are
integrated, rather than designed in isolation.
MAIN FRAMEWORK
The establishment of performance-based fire protection analysis and design method for large cross-section
road tunnel is under the background of performancebased code for fire protection design of building and
the prescriptive-based code for fire protection design
of tunnel. The whole performance of fire protection
for road tunnel should be comprehensively considered. By means of developing design fire scenarios,
the fire protection goals, such as provide life safety,
protect property and fireproof of tunnel structure, have
to be anticipated. Thereby, the fire protection level of
road tunnel could be synthetically assessed according
to design criteria and the corresponding fire protection
measures could be determined.
545
The construction process of the main framework

involves three steps: (1) Concept Establish; (2) System
Analysis; (3) System Design. Among these steps, step
of Concept Establish is to definite core need of the
method, step of SystemAnalysis is to set up total model
of the framework and step of System Design is to create
the implement system.
2.1
Module 2:
Information
Integration
Concept establish
In order to enable the method be applied in the practical fire protection engineering of large cross-section
road tunnel, the fundamental concept of the framework should be established on the basis of current fire
protection codes.
During the evaluation process of fire protection
feature for large cross-section road tunnel, the performance of corresponding fire protection facilities
could be confirmed. The traffic volume in the road
tunnel could be anticipated and the related fire protection management could be ascertained. In general
situation, the assumed conditions tally with practical project. As for some special situation, the further
research works have to be carried out.
2.2
Module 1:
Comprehensive
Evaluation
System analysis
Based on fundamental concept of framework establishment, facing qualified designer, authority having
jurisdiction, owner and management personnel involving fire protection of road tunnel, the basic framework
of fire protection method for large cross-section road
tunnel is constructed. The framework consists of three
modules, including Comprehensive Evaluation, Information Integration as well as Performance Prediction
respectively.
The framework of performance-based fire protection analysis and design method for large cross-section
is shown in Figure 1.
The main function of modules is as follows.
Module 1: Comprehensive Evaluation
This module is the key part of the framework,
including main functions such as:
1. Define object of fire protection design;
2. Establish design criteria of fire protection;
3. Evaluate comprehensively the fire protection
features.
Module 1 is appropriate for stakeholders of
road tunnel, including road tunnel owner, qualified
designer, authority having jurisdiction and insurer, etc.
Module 2: Information Integration
This module is the basic part of the framework,
including main functions such as:
1. Collect corresponding design materials according
to current prescriptive-based code. Determine different kinds of parameters for fire protection design
of road tunnel;
Fire Safety
Engineering
Module 3:
Performance
Prediction
Prescriptive-based Fire
Protection Design Code
Figure 1. Framework of performance-based fire protection analysis and design method for large cross-section road
tunnel.
2. Collect basic data of fire protection design. The

detailed design parameters included road tunnel
feature, vehicle variety, fire characteristic and fire
protection facility, etc.
3. Definite other comprehensive performance, including background information related with fire protection design.
Module 2 is applicable for qualified designers of
road tunnel.
Module 3: Performance Prediction
This module is the quantitative analysis part of the
framework, including main functions such as:
1. Define design fire scenario. According to statistical
data of tunnel fire and design goal of fire protection,
the design fire scenarios should be chosen.
2. Definite performance anticipation approach. The
analyses may be deterministic or probabilistic,
deterministic method would normally be used.
3. Predict system performance. Taking the whole fire
process as object of study, all components or subsystems which possibly affect the fire protection
feature of the road tunnel should be thoroughly
analyzed. As for performance evaluations, the components or subsystems of the fire safety system
as well as the interactions between the subsystems
would need to be considered. The subsystems that
may be concerned include: (1) fire initiation and
development; (2) spread, control and management
of smoke; (3) fire detection; (4) fire suppression;
(5) personnel behavior and egress; (6) passive fire
protection.
546
Module 3 is applied for professional assessment

personnel and researcher.
2.3
start
System design
In accordance with the function specificities determined by the three modules for fire protection analysis
and design method of large cross-section road tunnel,
the detailed content of corresponding implementation
structure constructed by System Design is presented
in Figure 2.
Define scope of large

cross-section road tunnel
Collect related
performance
information
Determine design
goal and objective
of fire protection
2.3.1
Define scope of large cross-section road

tunnel
The scope of large cross-section road tunnel is an
identification of the boundaries related with the
performance-based analysis or design, including: road
tunnel constraints, design and construction team organization, project schedules, applicable regulations and
other useful information to assist in comprehending
the scope definition.
Establish design criteria

of fire protection
2.3.2
Determine design goal and objective of

fire protection
Fire protection of large cross-section road tunnel generally has four interrelated fundamental fire safety
goals: (a) Provide life safety for the public and fire
fighters. Minimize fire-related injuries and prevent
undue loss of life. (b) Protect road tunnel. Minimize
damage to road tunnel from fire and fire protection
measures. (c) Provide for continuity of road tunnel
operations (i.e., protect the passage of vehicles). Minimize undue loss of operations due to fire-related
damage. (d) Limit the environmental impact of fire
and fire protection measures.
Once the fire protection goals have been established
and agreed to, the design objectives to meet the fire
protection goals must be defined. The design objective provides more detail than a fire protection goal,
and is often stated in terms meeting the requirements
of a specific code or standard provision (prescriptiveor performance-based), of a specific insurance-related
requirement, or in addition to a specific code, standard,
or insurance provision or requirement.
Determine design fire scenario
Develop trial design of

fire protection for tunnel
Evaluate
performance
of fire safety
system
Assess
evacuation of
personnel &
vehicles
Meets
criteria ?
No
Yes
2.3.3 Collect related performance information

There are many types of information that may
affect road tunnel design performance that should
be considered and collected, including characteristics such as: vehicle and personnel, location of the
tunnel, fire service, utilities, environmental considerations, tunnel management and security, economic
and social value of the tunnel, the tunnel delivery
process, applicable regulations, etc: It is quite important to identify the appropriate codes, regulations
and insurance requirements for the performance-based
analysis.
Select fire protection

strategies for tunnel
end
Figure 2. Fundamental content of performance-based fire
protection analysis and design method for large cross-section
road tunnel.
547
2.3.4 Establish design criteria of fire protection

The design criteria of fire protection for road tunnel in this paper are mainly concerned performance
criteria. Performance criteria are usually threshold values, ranges of threshold values, or distributions that are used to develop and evaluate trial
designs for a given design of large cross-section
road tunnel. Performance criteria may include temperatures of materials, gas temperatures, smoke concentration or obscuration levels, radiant flux levels,
and human response, decision, reaction, and movement times. For example, performance criteria may
include values for thermal radiation exposure (kW/m2 )
or gas (air) temperature. Other types of performance criteria include concentration of toxic gases
(ppm), distance of the smoke layer above the floor
(m), visibility(m), or other measurable or calculable
parameters.
2.3.5 Determine design fire scenario
In case the performance criteria have been established,
the qualified designer needs to focus on the development and analysis of design alternatives to meet
these criteria. The possible fire scenarios should be
considered are then filtered into selected design fire
scenarios. As design fire scenarios constructed, then
trial designs of road tunnel could be developed and
evaluated to determine whether the performance criteria will be successfully met by the trial design for a
given design fire scenario.
A fire scenario represents one of a set of fire conditions that are thought to be threatening to a certain
road tunnel. For a given fire scenario of road tunnel,
there are many factors that may affect fire development. These different factors may include: (a) form
of ignition source; (b) different items first ignited;
(c) ignition in different locations of a road tunnel;
(d) effects of tunnel geometry; (e) ventilation, whether
longitudinal ventilation or transverse ventilation, etc;
(f) form of intervention (i.e. personnel, sprinklers, the
fire department, etc.).
2.3.6
Develop trial design of fire protection

for tunnel
After the performance criteria have been established
and the design fire scenarios have been determined,
the trial design of the road tunnel should be developed.
Developing trial design may simply require selecting features similar to that of the prescriptive based
design option, but with enhanced capabilities or features. The design features being developed should
give consideration to the capabilities, reliability, costs
and maintenance. However, it should be considered
that the trial design would be evaluated. Trial design
can be evaluated on a system performance that relies
on an evaluation relative to established performance
criteria.
2.3.7 Evaluate trial design of fire protection

for tunnel
Evaluation is the process of determining if a trial
design meets the performance criteria during the
postulated design fires. The intent is to demonstrate
that in the design fire scenario, performance criteria
will not be exceeded.
If the trial design of the road tunnel is found successful, any remaining trial designs may be evaluated
as necessary. If one of trial designs is not found to be
successful, then it may be modified and re-tested, or
it may be dismissed. After the selected trial designs
have been tested, a final design should be selected
from among those found successful. If there are no
trial designs that are found successful, the qualified
designer should ensure that the trial designs considered
all possible mitigation strategies. If after considering
all possible mitigation strategies, there still are not
any trial designs that are found successful, the design
goal and objective of fire protection as well as the
performance criteria should be reexamined.
Many techniques can be used to evaluate the adequacy of a trial design for road tunnel in case of fires.
These techniques typically fit into two principal categories, probabilistic and deterministic. A deterministic
analysis examines the hazard posed by the potential
design fire scenarios independently. A probabilistic
evaluation uses risk analysis to identify consequences
of specific events and their respective likelihood.
The evaluation of a performance-based fire protection design for a large cross-section road tunnel is
a function of several factors. These factors include:
(a) complexity of road tunnel geometry; (b) level of
subsystem interaction; (c) type of performance criteria; (d) sensitivity of subsystem output to design
objectives; (e) absolute or comparative evaluation;
(f) knowledge level; (g) benefit versus cost; (h) expert
judgment and experience. The evaluation should also
account for known variations and uncertainties.
The main evaluation content of the trial design for
road tunnel may concern evaluating performance of
corresponding fire safety system as well as assessing
evacuation of personnel and vehicles, etc. It is necessary to point out that timelines could be valuable
tools in evaluation. Therefore, it may be essential to
determine the time of key events such as: (a) ignition;
(b) fire detected; (c) evacuation begins; (d) untenable conditions reached in road tunnel; (e) fire spreads
scope; (f) suppression begins; (g) failure of structural
elements; (h) fire extinguished.
2.3.8 Select fire protection strategies for tunnel
In term of performance-based fire protection design of
a large cross-section road tunnel, while the acceptable
trial designs are identified by the evaluation, they can
be considered for the selection of fire protection strategies. The choice of which acceptable trial design is
548
selected for the final design may be based on a variety

of different factors, including financial considerations,
timeliness of installation, system and material availability, ease of installation, maintenance and use, and
other factors.
Once the fire protection strategies are identified, the
relevant design documents need to be prepared. Proper
documentation will ensure that all stakeholders understand what is necessary for the strategies implementation, maintenance and continuity of the fire protection
design. The documentation should include the design
brief, a performance design report, detailed specifications and drawings, and a road tunnel operations and
maintenance manual.
3
CONCLUSION
It is evident that the framework of performancebased fire protection analysis and design method
for a large cross-section road tunnel offers a number of advantages over prescriptive-based design. It
provides a basis for development and selection of
alternative fire protection options based on the road
tunnels needs and results in a composite fire protection strategy in which all fire safety systems are
integrated, rather than designed in isolation. Hence,
such a comprehensive engineering approach often
provides more cost-effective fire protection solution

for the road tunnel. Apparently the intensive study
about the framework should be further conducted.
ACKNOWLEDGMENTS
REFERENCES
Eberl, G. 2001. The Tauern Tunnel Incident: What happened
and What has to be Learned, Proc. 4th Int. Conf. on Safety
in Road and Rail Tunnels, Madrid, Spain, 26th April
2001: 1730.
Han, X., Shen, Z.Y. et al. 2000. Study on a framework of
performance-based assessment method for fire safety of
large public building, Proceeding of the Sixth International Symposium on Structural Engineering for Young
Experts,Yunnan Science and Technology Press, Kunming,
China, August 2000: 1820.
Lacroix, H.D. 2001. The Mont Blanc Tunnel Fire: What Happened and What Has Been Learned, Proc.4th Int. Conf.
on Safety in Road and Rail Tunnels, Madrid, Spain, 26th
April 2001: 316.
Turner, S. 2001. St. GotthardTunnel Fire, New Civil Engineer,
1st Nov. 2001: 57.
549
Prediction of surface settlements induced by shield tunneling: An ANFIS

model
J. Hou, M.X. Zhang & M. Tu
Department of Civil Engineering, Shanghai University, Shanghai, P.R. China
ABSTRACT: A new method fuzzy system combination neural network was used to estimate ground surface
settlements. According to the measured data of Shanghai No.2 Subway, and considering various kinds of factors
synthetically, an ANFIS fuzzy neural network prediction model was built. Comparing with the prediction results
by other three kinds of methods, the validity of the ANFIS fuzzy neural network model was appraised. Confirmed
by the instance, ANFIS fuzzy neural network is valuable in predicting ground settlements induced by shield
tunneling.
INTRODUCTION
Shield tunneling has become one of the most popular methods used in the construction of urban tunnels,
such as rapid transit systems and large diameter underground pipelines. However, shield tunneling construction inevitably disturbs the ground and the original
stress field of soil, which in turn causes surface settlement that may yield damage of existing adjacent
structures and underground facilities. Therefore, it is
of significant importance for engineers to accurately
predict the surface settlement during the design and
construction stages.
Considering the complexity of the problem that
involves intricate geological makeup of the ground
composed of different materials with varying layer patterns plus different construction methods, it is apparent
that using classical method makes it difficult to provide
accurate predictions of ground settlement. In addition,
it is very difficult to determine input parameters representative of the mixed geological compositions in
a prediction model. Therefore, the analytical methods
relying on observed data are widely used in investigation of surface settlement. A hybrid intelligent system
called ANFIS(Jang 1993) (Adaptive-Network-Based
Fuzzy Inference System) combining the ability of a
neural network to fuzzy logic have the advantages of
both neural networks (e.g. learning abilities, optimization abilities, and connectionist structure) and fuzzy
systems (e.g. humanlike if-then rules, and ease of
incorporating expert knowledge and judgment available in linguistic terms). Such a hybrid intelligent
system holds much potential in prediction. In this
paper, a model based on ANFIS is proposed to predict
the ground settlement induced by shield tunneling.
Figure 1. Architecture of ANFIS.
DEFINITION OF THE MODEL
2.1 Brief description on the

Adaptive-Network-Based Fuzzy Inference
System (ANFIS)
Both artificial neural network and fuzzy logic are used
in ANFISs architecture (Chang & Lee 2003). The
ANFIS is one of the methods to organize the fuzzy
inference system with given input-output data pairs.
The ANFIS optimizes the parameters of consequent
part using least square method and those of premise
part using steepest descent method.
For simplicity, we assume the fuzzy inference system under consideration has two inputs (x and y) and
one output (z) in the present study. The corresponding
ANFIS architecture is shown in Figure 1.
Layer 1: In this layer x and y represent different
types of input parameters of the adaptive node i of an
adaptive function, and Ai is the linguistic label associated with this node function. Parameters in this layer
are referred to as premise parameters.
Layer 2: Every node in this layer is a fixed node
labeled which multiplies the incoming signals and
551
sends the product out. Each node output represents the

firing strength of a rule.
Layer 3: Every node in this layer is a fixed node
labeled N. The ith node calculates the ratio of the ith
rules firing strength to the sum of all rules firing
strengths.
Layer 4: Every node i in this layer is an adaptive
node. Parameters in this layer will be referred to as
consequent parameters.
Layer 5: The single node in this layer is a fixed
node labeled that computes the overall output as the
summation of all incoming signals.
The steepest descent method can be applied to find
the premise parameters and least square estimate can
he applied to optimize the consequent parameters.
Table 1.
Samples used for network training.
Serial
number X1
X2
X3
X4
X5
X6 y
1
2
3
...
81
11.65 24.70 31.60 34.35 34.85

1.50 7.65 24.00 36.40 42.29
6.80 52.10 72.57 89.67 98.75
...
3.75 14.85 38.20 36.85 39.70
Table 2.
Samples used for network validation testing.
Serial
number
X1
X2
X3
X4
X5
5
5
8
21.20
55.25
70.30
57.05
X6
2.2 Input and output parameters

A relatively too large surface settlement may induce
damage or even crash of existing structures. It is therefore very crucial to accurately predict the total surface
settlement (the maximum surface settlement after construction) of a point before the shield arrives at this
point. It is straightforward to set the total ground surface settlement to be an output variable in a prediction
model. In this paper, similar to many other researchers,
we choose the total surface settlement 5 meters ahead
of the shield face as an output parameter. However,
how to choose the input variables depends on the collection of monitored data. Systematic analyses of the
measured data show that the disturbed area by shield
tunneling of the original stress field of soil is approximately limited in a zone within 15 m from the working
point, and almost has no effect on the area out of this
zone. M. Karakuss measured data showed also that
the maximum settlement is over the tunnel centerline.
Considering these two aspects, we choose six parameters, i.e., five total ground surface settlements at the
points over the tunnel centerline (i.e., 0, 5, 10, 20 and
30 meters behind the tunnel face) and one of the shield
working parameters, the total number of the working
cycles of the day, as the inputs.
The investigation in this paper is based on the measured data in the shielding tunnel from Zhongshan Park
station to Longdong Road station in the Shanghai No. 2
Subway Tunnel Project. The total length of the tunnel
is 1624 m. The earth pressure balance (EPB) shield
was used in the project. The outer diameter of the EPB
machine is 6340 mm and the length is 6540 mm. The
tunnel shield tunneling was started in 18 July, 1997
and completed by 9 November, 1997, and the surface settlements were measured from 21August. Large
numbers of surface settlement markers were installed
to measure surface settlements during excavation. The
installation of surface settlement markers is described
in detail by Sun and Yuan. The soil profile around
the tunnel can also be found in the paper. The collected data have formed a database. The database not
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
5.20 20.30 26.75 23.20

9.95 8 40.95
0.05 1.07 0.35 2.45
4.95 12 62.45
13.07 54.42 82.62 96.60 25.45 11 89.87
10.00 27.25 63.00 81.10 91.60 11 53.50
14.45 51.30 57.85 74.85 90.52 10 54.50
4.75 12.65 28.55 38.11 46.20 4 57.00
6.80 52.10 72.57 89.67 98.75 8 83.15
3.20
9.75 2.60
1.42 2.27 11 99.57
4.17
7.32 7.32 17.00 44.90 7 41.10
14.05 16.15 14.39 22.1 34.50 8 47.22
0.95
2.85 16.25 54.45 47.65 12 76.42
3.30 15.55 33.30 45.30 52.00 6 53.90
1.70 15.60 30.25 47.05 55.40 11 47.30
2.95
4.95 13.75 44.15 86.40 10 61.30
13.25 43.75 76.30 74.85 68.52 10 56.73
3.75 22.02 47.77 66.90 72.75 13 88.15
4.57 18.75 46.40 68.75 47.15 7 79.02
4.00 11.65 34.05 47.65 77.40 8 61.30
0.72
6.92 18.37 43.37 51.65 10 79.60
1.90 12.20 30.40 49.65 39.30 11 71.95
only allows one to study the behavior of ground movements occurring during excavation, but also becomes
a useful source for developing predictive models for
the ground settlement. We randomly chose 81 in the
database experimental data (from August to October
in 1997, see Table 1, detailed list can be found in Tu
(2005)) as a training set and 20 data as a validating set
(data in November, see Table 2).
Here, X1 , X2 , X3 , X4 , X5 is total ground surface
settlements (unit: mm) at the points over the tunnel
centerline (i.e., 0, 5, 10, 20 and 30 meters behind the
tunnel face) separately. X6 is the total number of the
working cycles of the day.
2.3 Data pretreatment

To speeding calculation, we standardize the original
data to have a minimum of 0 and a maximum of 1 by
552
Table 3. The comparison of predicted surface settlement

and measured settlement.
120
Prediction
Experiment
100
80
60
40
20
0
6
12 18 24 30 36 42 48 54 60 66 72 78
serial number of measured point
Figure 2. The fitting map of the ANFIS output and the

measured data.
linear transformation. It is shown that such a transformation will make the calculation very efficient. The
algorithm of the linear transformation is as follows.
The original data of six input variables can be
expressed as:
We note that i = 1, 2, . . . P is the number of samples

and totally 81 samples are used in this paper to train
the model, i.e., P = 81.
We normalize each input datum as
Then the normalized input data can be expressed as
3.1 Text and indenting

The present model is realized via a Matlab package of ANFIS. The fitting map of the ANFIS output and the measured data was obtained (Fig. 2).
Serial
number
Measured
settlement
(mm)
Predicted
Settlement
(mm)
Absolute
error
(mm)
Relative
error
%
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
40.95
62.45
89.87
53.5
54.5
57
83.15
99.57
41.1
47.22
76.42
53.9
47.3
61.3
56.73
88.15
79.02
61.3
79.6
71.95
39.24
63.3
92.34
52.46
52.59
58.38
81.55
103.28
38.25
45.39
74.23
54.96
48.71
62.37
54.42
86.29
80.16
63.91
78.22
73.28
1.71
1.15
2.47
1.04
1.91
1.38
1.6
3.71
2.85
1.83
2.19
1.06
1.41
1.07
2.31
1.86
1.14
2.61
1.36
1.33
0.042
0.018
0.027
0.019
0.035
0.027
0.019
0.037
0.069
0.039
0.029
0.02
0.03
0.017
0.041
0.021
0.014
0.043
0.017
0.018
The comparison of predicted surface settlement and

measured settlement was also shown in Table 3.
They all show that all the predicted results from the
ANFIS model are in good agreement with measured
data.
Peck (1969) presented the first available method
for estimating the ground settlement due to tunneling
and excavation. In his method, charts were developed
based on the field data obtained from subway constructions at different places worldwide. The data points,
though scattered, revealed the shape of a Gaussian
distribution curve. The charts have been widely used
for estimating the transverse ground settlement profile
caused by tunneling and excavation. Theoretical models based on the combined pi-sigma approach (Gupta
& Rao 1994) and BP (Back Propagation) neural network are also used to predict the surface settlement.
Due to the limitation of the paper length, we do not
discuss them here.
Comparison between the above methods and the
ANFIS method is shown in Fig. 3.
It is found that the current prediction is in good
agreement with measurement. The maximum error
between the model and the testing data is not more than
7%. For comparison, we proposed also the predicted
results from two theoretical models based on the BP
(Back Propagation) neural network and the combined
pi-sigma approach (which are frequently used in predicting the surface settlement), and those from Pecks
equation (Fig. 4). We can see from the figure that
553
120
110
measured data
ANFIS
Combined pi-sigma neural network
BP package
Peck formula
This paper has shown the potential for applying fuzzy

neural networks to ground surface settlement analysis. five total ground surface settlements at the points
over the tunnel centerline (i.e., 0, 5, 10, 20 and 30
meters behind the tunnel face) and one of the shield
working parameters, the total number of the working
cycles of the day, as input parameters to predict the
surface settlement induced by tunnel shield tunneling.
The predicted results from the proposed model are in
good agreement with field observations and the maximum error between the model and the testing data is
not more than 7% in our example. Comparison with
some other predicting approach shows that the present
model is accurate, steady and efficient.
100
90
80
70
60
50
40
30
0
10
15
20
Figure 3. Comparison of predicted results with measured

ones.
ANFIS
Combined pi-sigma neural network
BP package
Peck formula
0.25
0.20
relative error
0.15
0.10
0.05
0.00
0.05
0.10
5
10
15
CONCLUSION
20
Figure 4. Comparison relative error from ANFIS to other

methods.
REFERENCES
Chang, K.H. & Lee, J.J. 2003.Adaptive Network-based Fuzzy
Inference System with Pruning. SICE Annual Conference
in Fukui, 46 August 2003. Japan: Fukui University.
Gupta, M.M. & Rao, D.H. 1994. On the principles of fuzzy
neural networks. Fuzzy Sets and Systems, 61(1):18.
Jang, J.S. 1993. ANFIS: Adaptive-Network-based Fuzzy
Inference Systems. IEEE Transactions on Systems, Man
and Cybernetics, 23(3): 665685.
Karakus, M. & Fowell, R.J. Effects of different tunnel face
advance excavation on the settlement by FEM. Tunnelling and Underground Space Technology, 2003, 18(5):
513523.
ground. Proceedings of 7th International Conference on
Soil Mechanics and Foundation Engineering, Mexico.
Sun, J & Yuan, J.R. Soil disturbance and ground movement
under shield tunnelling and its intelligent prediction by
using ANN technology. Chinese Journal of Geotechnical
Engineering, 2001, 23(3):261267. (in Chinese)
Tu M. The prediction of the ground settlements induced by
based on the Fuzzy Neural Network. Shanghai university
masters thesis. 2005. (in Chinese)
the BP neural network has the maximum error among

these methods; that the combined pi-sigma approach
underestimates but the Pecks formula overestimates
the surface settlement.
554
Experimental studies of a geological measuring system for tunnel with

ultrasonic transducer
D.H. Kim, U.Y. Kim, S.P. Lee & H.Y. Lee
GS Engineering & Construction, Seoul, Korea
J.S. Lee
Korea University, Seoul, Korea
ABSTRACT: Predicting ground conditions ahead of the tunnel face has been one of the most important
requirements of tunnel construction. This study investigated the development and application of a high resolution
ultrasonic wave imaging system, which captures the multiple reflections of ultrasonic waves at the interface, to
detect discontinuities at laboratory scale rock mass model. Ultrasonic wave reflection imaging based on A- and
B-modes was obtained through stacking, signal compensation, demodulation, and display. Experiments were
carried out by using horizontal scanning and rotational scanning. Experimental studies showed that the rotational
scanning method was able to identify horizontal and inclined discontinuities and the cavity on the plaster block
at a fixed location. Furthermore, two discontinuities including horizontal and inclined discontinuity planes were
detected. The rotating scanning technique produced images similar to those obtained by the typical horizontal
scanning technique. This paper contains basic theories about the ultrasonic transducer and several experimental
application results. The full-scale field application and other application will be scheduled in the future.
INTRODUCTION
With the rapid growth of the worlds population and

economics, increasing number of tunnels have been
constructed to provide better transport links. Safe
and economic tunneling has always been a challenging topic in tunnel construction because of complex
ground conditions like faults, fractures, caverns, and
wet layers. Therefore, the prediction of ground conditions ahead of tunnel face is considered one of the
important requirements.
Many techniques to predict the ground condition ahead of the advancing tunnel face have been
developed, improved and widely applied in tunnel
construction projects. Three types of techniques are
representives for tunnel ground condition predicting:
coring, displacement monitoring and analysis, and
wave-based non-destructive evaluations(Figure 1(a)).
In this study, a new high resolution seismic technique based on rotating scanning equipment is introduced to detect geological discontinuities, as shown
in Figure 1(b). This technique uses the received time
series which captures multiple reflections due to
the impedance mismatch at a geological interface.
The source and receiver transducers are located on the
tunnel face at one fixed point. As a preliminary study
of this technique, this paper describes experimental
studies carried out on lab-scaled models. This paper
Figure 1. Seismic method for predicting tunnel face:

(a) TSP method; (b) Rotating scanning method.
presents basic theories about the ultrasonic transducer

including transducer frequency characteristics, transducer beam characteristics, and coupling layer characteristics, experimental setup and signal processing,
application examples, discussion, and conclusions.
2
ULTRASONIC TRANSDUCER
Ultrasonic wave reflection imaging system may be an

economical and effective tool for the forecast of the
ground condition ahead of the tunnel face. A transducer refers to any device that can convert an electrical
signal into a mechanical energy and vice versa. In
ultrasonic reflection imaging, a piezoelectric leadzirconate-titanate (PZT) type transducer is generally
used to as a source and detector (Wells, 1977). The
selection of the transducer is the most important factor
555
in ultrasonic reflection imaging because the transducer

determines the image resolution and skin depth.
2.1 Transducer frequency characteristics
The choice of transducer frequency in ultrasonic
reflection imaging is the result of a compromise
between the resolution (lateral and axial) requirement
and the acquirement of satisfactory beam penetration
for the imaging of the part of interest. Lateral resolution refers to the ability to distinguish two closely
spaced reflectors, which are positioned perpendicular
to the axis of the ultrasound beam. Lateral resolution
is most closely related to the transducer beamwidth.
Axial resolution refers to the minimum reflector spacing along the axis of an ultrasonic beam that results in
separate, distinguishable echoes on the display.
For a given frequency, the shorter the pulse duration, the wider the frequency bandwidth. If a shorter
pulse duration is used, high resolution can be attained,
but it lowers the sensitivity and skin depth. But sensitivity can be improved by either increasing the energy
of the transmitter or by amplifying the captured signals at the receiver. Note, skin depth is the ability to
detect an anomaly at a given depth and it depends on
the amplitude of the reflected signal. When the resolution increases with increasing frequency, the skin
depth decreases (Lee and Santamarina 2005).
2.2 Transducer beam characteristics
Near Field and Far Field. The transducer beam is characterized as near field and far field, which are sketched
in Figure 2. In the near field, which is called the Fresnel
zone, wave amplitude fluctuates. Note the beams are
almost parallel rather than divergent in the near field.
The near field length NFL or Fresnel zone length is
dependent on the transducer radius r and wavelength
(Krautkramer and Krautkramer, 1990; Rose, 1999).
where f is the frequency, and V is the ultrasonic wave

velocity of the medium. Therefore, the near field
length NFL increases with the increase of the transducer radius r and/or the increase of the frequency f.
The zone beyond the near field is the far field, which is
also called the Fraunhofer zone. The divergence angle
depends on the directivity.
Directivity. The far field divergence angle depends
on the wavelength and radius r (or diameter d) of the
transducer (Zagzebski, 1996), as shown in Figure 2.
The divergence angle should be made small to

increase lateral resolution. Higher frequency and
larger diameter render higher lateral resolution (far
field is assumed).
Figure 2. Beam characteristics of transducer.
Figure 3. Transducer type and focal length: (a) Schematic

drawing of transducer; (b) Focal length.
The directivity of the transducer is affected by the

velocity of the test medium, and the frequency and size
of the transducer. High directivity yields higher resolution. The directivity of an ultrasonic beam may be
altered by focusing the beam: 1) a curved PZT element
instead of a plane PZT element is used and 2) a lens is
attached in front of the plane PZT element (Schmerr,
1998), as shown in Figure 3. Because the wave velocity in lens is greater than that in water, the shape of the
lens will be concave. The focal length F is defined as
the distance from the point on the curved surface on
the central axis to the midpoint of the region of convergence. Note, if the lens is used, the focal zone instead
of the focal point appears, as shown in Figure 3(b). The
focal length is dependent on the radius of curvature R,
the radius of the lens a, and the aperture angle 2 (see
Figure 3).
For the investigation of the directivity, two types
of transducers are used: focal type transducer (PanametricsA3441) and non-focal transducer (Panametrics
V318). Both are high damping immersion transducers
having 500 kHz in resonant frequency and 19 mm in
diameter. Directivities were investigated by measuring the amplitude of the signal at fixed axial distances
(50 mm, 100 mm and 150 mm) and the lateral offsets
of the transducers, as shown in Figure 4(a). A focal
556
Figure 4. Directivity: (a) Test procedures; (b) Focal type

transducer; (c) Non-focal transducer; (d) Lens focal type
transducer.
Figure 5. Coupling layer effects: (a) Vacuum grease-not

adhered tightly; (b)Vacuum grease-adhered tightly; (c) Water.
type transducer has good directivity, as shown in Figure 4(b). Whereas a non-focal type transducer diverges
much of the energy, as shown in Figure 4(c), a nonfocal type transducer with a concave lens (lens focal
type transducer) improves the directivity dramatically,
as shown in Figure 4(d).
Coupling layer characteristics. The coupling layer,
which is an agent between transducer and medium,
should minimize the reflection from the surface of the
medium tested to maximize the energy transferred to
the medium. Vacuum grease and water were examined
as agents in this study. The test medium was a plaster block of 300 mm in height, 300 mm in width and
150 mm in thickness. The minor energy was reflected
with vacuum grease when the transducers were lightly
placed on the top of the vacuum grease (it is not
adhered tightly), as shown in Figure 5(a). However,
when the transducers were tightly adhered, the amplitude of the reflected signal increased, as shown in
Figure 5(b). Note, the amplitude of the directly transmitted wave, which propagates from the source to the
receiver through the tested medium, also increases.
Furthermore, it is very difficult to maintain the constant contact between the transducer and the vacuum
grease during the scanning. When water was used as
a coupling layer, it produced a relatively high amplitude reflection as shown in Figure 5(c). In addition, the
directly transmitted wave can be effectively shielded
by using several layers of aluminum foil. Therefore,
water was selected as the coupling layer for scanning
in this study.
Figure 6. Signal processing: (a) Measured signal after

stacking; (b) logsig function(window function) for time gain
compensation; (c) Compensated signal; (d) Rectification;
(e) Moving averaging; (f) Rejection.
EXPERIMENTAL SETUP AND SIGNAL

PROCESSING
3.1 Experimental setup

A Pulser (JSR, DPR300) was used to generate the
ultrasonic waves through the transducer (Panametrics
A3441). Two transducers were used as the source and
receiver, respectively. The reflected signals due to the
impedance mismatch at the interface were detected
and converted into electrical signal by the receiver
transducer. The electrical signal was fed through an
amplifier (Krohn-Hite 3945: frequency range from
170 Hz to 26.5 MHz) because the amplitude of the signals measured by the receiver was generally too low to
identify the meaningful reflection. The amplified signal was digitized in the oscilloscope (Agilent 54624A
or National Instruments PXI-5112).
3.2 Signal processing
Signal processing for ultrasonic imaging consists of
stacking, signal compensation, demodulation including rectification and smoothing, rejection, and display
(Zagzebski 1996).
Stacking. Signals were captured after stacking,
which is the most effective signal processing technique
for removing the high frequency, uncorrelated noise.
The stacking means averaging multiple signals. The
1024 signals were averaged to obtain the signal trace,
as shown Figure 6(a).
557
Signal compensation. The amplitude of the

reflected signals at the receiver generally decreases
as the distance of the interfaces, which the waves
are reflected from, increases due to the geometrical
spreading under identical impedance mismatch. This
attenuation may be compensated by using time gain
compensation (TGC), which increases the amplitude
with time. Thus, the amplification factor is higher at
a longer distance than at a shorter distance from the
receiver. In this study, the adoptedTGC is a log sigmoid
(logsig) transfer function, as shown in Figure 6(b). The
compensated signal is obtained through point-by-point
multiplication between the original signal and logsig
function. Figure 6(c) shows the compensated signal
from the original signal. The amplitude of the first
and second reflections decreases.
Demodulation. Demodulation includes rectification and smoothing. Rectification is an inversion of
negative components. Thus, the signal has only positive values, as shown in Figure 6(d). Smoothing
(moving average) was carried out by using the kernel of [1/20, 2/20, 4/20, 6/20, 4/20, 2/20, 1/20] in this
study. The signal after the smoothing process is shown
in Figure 6(e).
Rejection. Rejection is an elimination process,
which removes the signal whose amplitude is less than
threshold value. Thus, rejection removes noises and
low amplitude signals. In this study, the threshold value
was about 20% of the maximum amplitude. The signal
after rejection is shown in Figure 6(f).
Display. Two modes were used to display the ultrasonic reflection imaging in this study. First, the amplitude mode (A-mode), which represents the amplitude
of the signal after rejection versus time, was used. In
the A-mode, the first arrival time of the reflected signal
can be easily determined. Note the height of the trace
in the A-mode is the amplitude. Second, the amplitude of each signal was converted to brightness, which
is proportional to the amplitude of the reflected signal. The brightness versus time is the brightness mode
(B-mode).
4 APPLICATION EXAMPLES
Several unique applications of ultrasonic wave monitoring and imaging were explored by using several
small scale plaster blocks and one large scale plaster
block.
4.1
Rotating scanning test for the small scale

plaster blocks
Three kinds of specimens were used for the rotating

scanning tests: an intact plaster block, a plaster block
with an inclined crack, and a plaster block with a cavity.
The dimensions of the plaster block were 300 mm in
height, 300 mm in width and 150 mm in thickness. The
middle of the top surface of the plaster block was dug
Figure 7. Rotating scanning test for an intact plaster block:

(a) Test setup; (b) A-mode image; (c) B-mode image.
Figure 8. Rotating scanning test for a plaster with crack:

(a) Test setup; (b) A-mode image; (c) B-mode image.
to form the concave shape for the rotational scanning,

as shown in Figures 79. Water, which was filled in
the concave part of the plaster block, was used as the
coupling layer.
Intact Plaster Block. The rotational scanning tests
on the intact block were carried out, as shown in Figure 8(a). The rotational scanning test on an intact plaster block clearly shows the interface at the bottom of
the plaster block as shown in Figures 7(b) and (c). The
brightest part in the B-mode occurred in the middle of
the bottom. Because the reflected signal at the bottom
of the plaster block was trapped in the water coupling
layer, the signals with the intermediate amplitude were
also detected below the bottom of the plaster.
Plaster Block with an Inclined Crack. After the
intact plaster block was cut into two parts in the
inclined direction as shown in Figure 8(a), the two
parts were filled with thin vacuum grease to maximize the energy transmitted through the crack. The
result of the rotational scanning test is shown in Figures 8(b) and (c). Two strong reflected signals were
detected from the inclined crack and from the bottom
of the block, respectively. The first strong reflection
occurred at the inclined crack. Note the inclination of
the crack was observed in the B-mode. The amplitude
or the brightness of the reflected signals increases as
the angle of incidence of the transducer beam becomes
zero degree with the normal to the crack (right side).
Note the brightest part in the B-mode corresponds to
the strongest reflection from the discontinuity, which
is perpendicular or normal to the transducer. Therefore, the brightest section can be used to determine
the angle of the inclined discontinuity. In addition,
because the reflected signal at the inclined crack was
also trapped in the water coupling layer, the multiple
reflections were observed. Note the inclination in the
558
Figure 9. Rotating scanning test for a plaster cavity: (a) Test

setup; (b) A-mode image; (c) B-mode image.
Figure 11. Effect of inclination angle: (a) Real discontinuity; (b) Ultrasonic reflection image.
Figure 10. Rotating scanning test for a large scale plaster

block: (a) Test setup; (b) B-mode image.
multiple reflections was identical with that of the first

reflection, which came from the inclined crack. The
second strong reflection was measured at the middle
of the bottom of the plaster block because the angle of
the incident wave became zero with the normal to the
bottom of the plaster block.
Plaster Block with a Cavity. A hole (Diameter =
30 mm) was drilled on the intact block, as shown in
Figure 9(a). The results of the rotational scanning test
are plotted in Figures 9(b) and (c). Strong signals were
also reflected from the bottom. The estimated diameter
of the cavity from the B-mode was about 43 mm. Note,
the difference between the real size (30 mm) and the
estimated size (43 mm) results from the divergence of
the beam, Fresnells ellipse, and the transducer size.
Although the inclined crack produced a continuous
reflection line as shown in Figure 8(c), the cavity only
yielded the reflections in a limited range. Furthermore,
the amplitude of the image from the cavity decreased
when the angle of the incident waves diverted from the
normal to the cavity.
4.2
Rotating scanning test for a large scale plaster

block
The dimensions of the large scale plaster block were

1000 mm in height, 1200 mm in width and 150 mm
in thickness. For the rotational scanning test, the left
part of the top surface of the plaster block was dug
to form the concave shape, and water was filled,
as shown in Figure 10. The rotational interval was
0.5 degree, which corresponded to 8.7 mm horizontal
displacement at the bottom of the large scale plaster
block. To simulate the multiple discontinuities, two

discontinuities were prefabricated: one was horizontal, and the other was inclined. The first horizontal
crack whose gap and length were about 10 mm and
250 mm was filled with very weak and low viscous
plaster paste for partial transmission and reflection
of the ultrasonic waves at this discontinuity. The second long and inclined crack was filled with vacuum
grease. B-mode is represented in Figure 10(b), that
shows clear reflections from the first horizontal crack
and the second inclined crack. In addition, the bottom
of the plaster block could be identified, even though it
was not clear.
5
DISCUSSION
5.1 Inclination angle of images

The inclination angle of the reflected image from
the crack was flatter than that from a real crack, as
shown in Figure 8. The relationship of the inclination angle between the reflected image and the real
discontinuity is
where, p1 and p2 are the depths of the discontinuity at

two points separated by the horizontal distances (see
Figure 11), and note, p1 in the reflected image is the
travel distance based on the travel time and velocity of the ultrasonic wave. The inclination angle of
the discontinuity estimated by Equation (4) may be
confirmed by the strongest brightness in the B-mode
(see Figure 8).
5.2 Horizontal scanning versus rotating scanning
After a paraffin block was installed underwater, typical
horizontal scanning (see details in Lee and Santamarina 2005) and rotating scanning tests were carried out.
The experimental setups are shown in Figure 12(a).
The thickness of the paraffin was 35 mm. For the
typical horizontal scanning test, the ultrasonic waves
559
when the incident waves diverges from the normal to

the interface.
6
Figure 12. Comparison between rotating scanning and horizontal scanning tests: (a) Test setup; (b) A-mode image; (c)
B-mode image.
reached perpendicularly to the interface. That is, the

angle of incidence of the transducer beam was always
zero degree with the normal to the interface. Thus,
only reflection and transmission occurred at the interface. The amplitudes of the reflected and transmitted
waves depend on the impedance mismatch. The scanning interval was set to 1 mm to avoid spatial aliasing
(wavelength in water is = 3 mm because the velocity of the ultrasonic waves in water is about 1500 m/s
and the frequency of the transducer is 500 kHz). Note
no migration processing was required because high
directivity transducers were used (Figure 4). Horizontal scanning could clearly detect the horizontal
paraffin wax layer in water. Furthermore, the bottom
of the water box was also clearly seen, as shown in
Figures 12(b) and (c).
For the rotating scanning test, the ultrasonic wave
may reach non-perpendicularly to the interface, and
therefore, a mode conversion may occur (see details
in Richart et al. 1970; Aki and Richards 1980). As the
angle of incidence of the transducer beam increases
from zero degree as shown in Figure 12(a), the arrival
time of the reflected wave increases, the amplitude of
the reflected waves decreases and finally diminishes,
as shown in Figures 12(b) and (c), for the rotating scanning test. However, B-mode obtained by the horizontal
scanning and by rotation scanning is almost identical
due to the size of the transducer (19 mm in diameter),
divergence (see Figure 4), and Fresnells ellipse. Note
images obtained by horizontal scanning and rotating
scanning techniques are related to the lateral resolution. Thus, the reflected waves are still measured even
CONCLUSIONS
The design and application of ultrasonic wave reflection imaging were documented in this study for the
detection of the discontinuity planes or cavities in
laboratory scale rock models. The signal processing,
which includes stacking, signal compensation, demodulation, rectification, smoothing, and rejection, were
carried out to produce the amplitude and brightness
modes (As and B-modes). The main observations of
this study follow:
Although vacuum grease transmits more energy to
the medium tested, water is recommended as the coupling layer for the horizontal and rotating scanning
techniques because constant contact area is maintained
in water and the directly transmitted wave between
transducers may be effectively removed.
The discontinuities of the plaster block, including horizontal and inclined cracks, and the cavity
were clearly detected by using the new rotating scanning technique. While the horizontal and inclined
cracks yielded continuous reflections, the cavity produced reflections at a limited zone. B-mode is more
appropriate for the detection of discontinuities.
The paraffin wax underwater was clearly detected
by the typical horizontal scanning and rotating scanning techniques. Furthermore, the two techniques
produced almost identical images. However, cautions
are required for the analysis of results obtained by the
rotating scanning technique.
The angle of the inclination obtained by the rotating
scanning technique may give the angle of the original
inclined crack through the comparisons of brightness
and through a simple calculation based on geometry.
REFERENCES
Aki, K. & Richards, P.G. 1980. Quantitative Seismology
Theory and Methods Vol. 1 and 2: 932. Freeman Company,
San Francisco.
Gomm, T.J. & Mauseth, J.A. 1999. State of the Technology: Ultrasonic Tomography, Materials Evaluation Vol.
57: 737755.
Krautkramer, J. & Krautkramer, H. 1990. Ultrasonic Testing
of Materials: 677. Springer-verlag, London.
Lee, J.S. & Santanmarina, J.C. 2005. P-Wave Reflection
Imaging. Geotechnical Testing Journal Vol. 28 : 197206.
Richart, F.E., Hall, J.R. & Woods, R.D. 1970. Vibrations of
Soils and Foundations: 414. Prentice-Hall, USA.
Schmmer, L.W. Jr. 1998. Fundamentals of Ultrasonic Nondestructive Evaluation A Modeling Approach: 559 Plenum
Press.
Wells, P.N.T. 1977. Biomedical Ultrasonics: 635. Academic
Press, London.
Zagzebski, J.A. 1996. Essentials of Ultrasound Physics: 220.
Mosby, Inc., Missouri.
560
Performance review of a pipe jacking project in Hong Kong

T.S.K. Lam
Geotechnical Engineering Office, Civil Engineering and Development Department, Government of
the Hong Kong Special Administrative Region
ABSTRACT: The pipe jacking method is commonly used in Hong Kong for construction of underground
cable duct crossings and stormwater drains. The method minimizes the disturbance to or interference with the
activities and facilities on the ground surface. In this paper, details of a pipe jacking project completed recently in
Hong Kong, involving use of a pressurized slurry tunnel boring machine to form a 222 m long, 1.95 m diameter
cable tunnel, are described. Results of the performance review carried out on completion of the project are also
presented.
INTRODUCTION
A cable duct crossing was constructed in an urban

area at Kowloon West of Hong Kong. The cable duct
crossing had to traverse a highway and some railway tracks. Conventional open cut excavation method
was not used because of the disturbance that could
be caused to the facilities on the ground surface. Pipe
jacking method was used instead. The cable duct crossing was constructed in fill comprising loose to medium
dense, silty coarse sand, and a pressurized slurry tunnel
boring machine (TBM) was selected for the project.
In this paper, details of the project are described. A
performance review was carried out on completion of
the project. The key construction aspects, monitoring
data obtained at the site during construction, impact
on sensitive features in and surrounding the site and
a summary of the observations and decisions made
during construction are presented. The information
given in this paper is obtained during auditing of the
project by the Geotechnical Engineering Office (GEO)
of the Civil Engineering and Development Department
(CEDD) of the Government of the Hong Kong Special Administrative Region. The auditing was carried
out to exercise geotechnical control in the interest of
public safety.
PROJECT DESCRIPTION
2.1 Details
The site is located near West Kowloon Highway at
Kowloon West (see Figure 1), Kowloon.
The project involves construction of a 222 m long
1.95 m diameter tunnel to serve as cable duct crossing
Figure 1. Location plan.
across a highway, the MTR Tung Chung line, the Airport Express Line and a public park. Pipe jacking
method was used to form the 2 m diameter pipe opening at 8 to 9 m depth in fill, the properties of which are
shown in Figure 8. The groundwater level was measured at about 2.0 m below ground.Two working pits of
14 m long 4 m wide 10 m deep were constructed
at both ends of the cable duct crossing for the tunneling
operation.
2.2
Geotechnical aspects of the tunnel works
At the design stage, plans and supporting documentation of the geotechnical design of the pipe jacking
561
Figure 3. Close-up view of settlement monitoring prism.
Figure 2. Locations of monitoring prisms on the railway

tracks.
works, a geotechnical risk assessment and a risk

mitigation plan were prepared by the designer.
The most sensitive features affected by the works
were the railway tracks and the underground utilities, including drainage pipes and sewers, close to the
alignment of the pipe jacking.
Prior to the commencement of works, a condition
survey of the existing road and the structures within
25 m of the alignment of the pipe jacking and a CCTV
survey of the existing drainage pipes and sewers within
20 m of the works were carried out. Another CCTV
survey was also carried out on completion of the
works.
An instrumentation scheme consisting of 168 settlement markers for the road surface, 123 settlement
monitoring prisms for the railway tracks and four
piezometers for groundwater level was adopted. Locations of the settlement markers, settlement monitoring
prisms and piezometers are shown in Figures 1 and 2.
The settlement markers were installed adjacent to
the jacking pit and receiving pit and on the road surface along the alignment of the pipe jacking. Out of the
123 settlement monitoring prisms installed, 32 were
installed on the Tung Chung Line (up) (TCL up) (on
both rails), 30 on the Airport Express Line (up) (AEL
up), 31 on the Tung Chung Line (down) (TCL down)
and 30 on the Airport Express Line (down) (AEL
down) (see Figure 2).
One piezometer each was installed close to the jacking pit (VBH1) and receiving pit (VBH2) and two
piezometers were installed along the alignment of the

pipe jacking (PZ1 and PZ2) (see Figure 1).
A monitoring plan including monitoring of settlement, groundwater level and vibration, and a site
supervision plan for the works were prepared. The
method statement for the works and details of the pressurized slurry machine selected were also submitted to
the relevant departments including the GEO for review.
The stakeholders affected by the project were also
notified and consulted by the designer. For the railway
tracks, trigger levels of alert, action and alarm
(or AAA) of 12 mm, 16 mm and 20 mm respectively
for settlement and 1 in 1500, 1 in 1250 and 1 in 1000
respectively for angular distortion were set and agreed
by the MTR Corporation Limited. At alert level,
readings would be reviewed and plans for remedial
measures and contingency actions would be prepared.
At action level, the planned remedial measures would
be implemented and the works could only continue if
the remedial measures taken were effective. The alert
and action levels would also be revised if necessary.
At alarm level, all works would be stopped or contingency actions taken and the design, construction
method and the planned remedial measures would be
reviewed. A maximum tolerable track settlement of
25 mm was allowed.
AnAutomatic Deformation Monitoring System was
used to take readings for monitoring the railway track
settlement. Three CYCLOPS theodolites were positioned on opposite sides along the tracks, and the track
settlement was monitored on a daily basis in real time.
The data taken were transferred to a computer in the
site office which were then processed and posted onto
the internet for real time monitoring by relevant parties
(Figures 3 to 5).
2.3 Construction
A RASA DHL1650 pressurized slurry TBM was
used (Figure 6). The diameter of the shield body
562
Figure 6. Pressurized slurry TBM.
Figure 4. Positions of settlement monitoring prisms on the

track.
Figure 7. Dimensions of TBM and concrete pipe.
Figure 5. A CYCLOPS theodolite.
and the equipment tube of this TBM is 2,000 mm

and 1,990 mm respectively. The excavated diameter is
2,040 mm. The pipe installed has an outside diameter of 1,950 mm and an internal diameter of 1,650 mm
(Figure 7).
The construction method involved use of a pressurized slurry system. The slurry support pressure at
the excavation face was set to balance the ground and
groundwater pressure as indicated in the control panel
of the TBM. The TBM operator checked the pressure
gauge to control the pressure at the excavation face and
the slurry-discharge pressure gauge to control the circulation pressure. The typical groundwater pressure at
the excavation face was 50 to 70 kPa with an average
value of 60 kPa at the tunnel axis level. To balance
the ground and groundwater pressure, a slurry pressure of 10 to 20 kPa above the water pressure, which is
60 to 90 kPa, was applied at the excavation face for the
operation (Figure 8). The slurry pressure applied to the
Figure 8. Typical groundwater pressure.
excavation face was controlled by the pressure control

valve of a by-pass unit placed in the jacking pit.
During operation, a lubricant consisting of water,
bentonite and mineral oil was injected into the annulus around the pipes to fill the voids around the pipes
to reduce soil movement. On completion of the pipe
jacking, the cavities around the pipes were grouted. A
cement grout with a minimum compressive strength
of 20 MPa was used.
563
3
3.1
PERFORMANCE REVIEW
Control of amount of materials excavated
In a pressurized slurry system, the ground was

mechanically excavated while the excavated face was
stabilized and supported by slurry. The excavated
materials were removed by slurry transport. To check
if voids had been formed in the soil above the tunnel
during excavation, the amount of materials excavated
was monitored during the tunneling operation. A desander was used to separate the solids from the slurry.
The solids were mainly fine sand which was later used
for backfilling the pits. The weight of the soil excavated from the jacking pit, receiving pit and the tunnel
was worked out to be 2,262 tonnes before the works
commenced. During the works, 593 tonnes of sand and
1,461 tonnes of finer soil were disposed of. The total
amount of soil disposed was therefore 2,054 tonnes
which was less than the calculated value. Although
limited by the accuracy of measurement, this provided
a rough indication that no significant voids had been
formed in the ground above the tunnel. No boulders
were encountered and no cutter had to be replaced for
this project.
3.2
Figure 9. Excavation sequence of pipe jacking (a) and settlement profiles at tracks TCL Up, AEL Up, AEL Down and
TCL Down (b).
Settlement of railway tracks
In this project, the most sensitive features are the

railway tracks.
The excavation sequence of the pipe jacking works
in terms of the distance from the railway tracks is
shown in Figure 9. Pipe jacking commenced on 29
March 2007. Fourteen days after commencement of
the pipe jacking (12 April 2007), the TBM reached the
TCL Up track and 5 more days later (17 April 2007),
it went past through all the tracks. Included in the figure is also the track settlement in response to the pipe
jacking.
The tracks settled as soon as the pipe was jacked
past and away from the section. At the TCL Up track,
the track settled by 12 mm when the pipe was installed
at a distance of 23.5 m away from the section. Similar
rate of settlement was noted at tracks AEL Up, AEL
Down and TCL Down except in those cases, the track
settlement was less. On completion of the pipe jacking,
maximum settlement of tracks TCL Up, AEL Up, AEL
Down andAEL Down were 22 mm, 15 mm, 10 mm and
3 mm respectively which are less than the maximum
tolerable track settlement of 25 mm.
In pipe jacking, the whole length of the pipe
is jacked and moved forward, resulting in disturbance to the soil around the whole length of the
pipe. With the cutter head about 90 mm larger than
the pipe to be jacked (see Figure 7), it is possible that voids would form around the pipe which
would result in ground movement, and hence settlement along the length of the pipe even after excavation
Figure 10. Settlement trough at track TCL Up.
has been carried out to some distance away from the

tracks.
The settlement troughs at tracks TCL Up, AEL Up,
AEL Down and TCL Down are shown in Figures 10
to 13 respectively. The settlement troughs are almost
symmetrical in these cases.
The maximum settlement on the pavement was
15 mm which was recorded by settlement point SC19
adjacent to TCL Up track (Chainage 29362).
Given the ground conditions and the method of construction as described above, a maximum volume loss
of 3.6% was calculated (see Figure 14).
The maximum groundwater drawdown recorded by
the piezometers at the jacking pit and receiving pit
(VBH2 and VBH1) was 1.1 m, which was only slightly
above the permissible value of 1 m. At PZ1 location,
564
Figure 11. Settlement trough at track AEL Up.
Figure 12. Settlement trough at track AEL Down.
Figure 14. Properties of error function to represent the

settlement trough above a tunnel (after Peck, 1969).
the maximum drawdown recorded was 0.7 m, and at

PZ2 location, no significant amount of groundwater
movement was observed.
No undue settlement was recorded at the drainage
pipes, sewers and the structures nearby.
the meeting, it was decided that for the remaining

works, an alternative type of lubricant, in lieu of a mixture of water, bentonite and mineral oil, was used to fill
the gap between the pipe and the TBM over-break. This
lubricant formed a solid substance once injected into
the gap and it helped to reduce settlement of the overburden fill caused by closure of the overbreak voids
around the pipe. The density of the bentonite slurry
was also increased. The advance speed of the TBM was
reduced from 9 m per night shift to about 6 m per night
shift. With the actions taken, the rate of increase of
track settlement was reduced and the final settlement
was contained to within the permissible value.
3.3 Actions taken at action level during

construction
Figure 13. Settlement trough at track TCL Down.
Twelve one days after commencement of the pipe

jacking (19 April 2007), a settlement of 15 mm was
recorded at the TCL Up track which was close to the
action level. An urgent meeting was held among the
client, designer and contractor on that night. After
CONCLUSIONS AND RECOMMENDATIONS
The following conclusions and recommendations can

be made from the project:
1. The use of pipe jacking method was effective in
forming a 222 m long, 1.95 m diameter cable tunnel
565
at 8 to 9 m depth. The pressurized slurry TBM used

is appropriate for the type of ground conditions
encountered.
2. It is necessary to apply appropriate geotechnical
control measures to this type of project affecting
sensitive features in the interest of public safety.
With the geotechnical control measures applied, the
geotechnical risk is assessed and the potential hazards are identified early. Any undue settlement or
irregularities observed during construction can be
detected promptly and appropriate remedial measures, such as use of a different annulus filler to suit
the actual ground conditions encountered, change
of the advance speed of TBM, etc. can be taken to
prevent catastrophic failure from happening.
3. It is important to control the amount of materials excavated from the tunnel opening to prevent
significant ground loss. This could be achieved by
comparing the excavated volume with the theoretical excavation volume to check if significant voids
have been formed in the soil above the tunnel and
checking the amount of filler/grout used to fill the
voids around the pipes. In this project, only the
total amount of materials excavated from the tunnel opening is obtained. A better method should be
devised for future similar operation measuring the
amount of materials excavated for each section of
the tunnel excavation and checking the amount of
filler/grout used to fill the voids around the pipes.
ACKNOWLEDGEMENT
The author would like to thank the Director of the
Civil Engineering and Development Department and
the Head of the Geotechnical Engineering Office for
the permission to publish this paper. Comments on the
paper from Mr Joe B.N. Leung and Dr Richard Pang
are gratefully acknowledged.
The author would also like to thank the Employer
of the Project CLP Power Hong Kong Limited, the
Designer Black & Veatch Hong Kong Limited and
the Main Contractor Kum Shing (K.F.) Construction
Co. Limited for the co-operation to exercise geotechnical control in the interest of public safety in the
project and the permission to publish the technical
data/details given in the paper. The views given in the
paper reflect only those of the author and not of the
above companies.
REFERENCES
ground. Proceedings of the 7th International Conference on Soil Mechanics and Foundation Engineering,
State-of-the-Art Volume: 225290.
566
Geotechnical control of a major railway project involving tunnel works

in Hong Kong
W. Lee, S.S. Chung, K.J. Roberts & P.L.R. Pang
Geotechnical Engineering Office, Civil Engineering and Development Department, Hong Kong Special
Administration Region
ABSTRACT: This paper describes the role of the Geotechnical Engineering Office, Civil Engineering and
Development Department of the Hong Kong Special Administration Region (HKSAR) Government in exercising
geotechnical control for a major railway project in Hong Kong. It highlights the geotechnical aspects of the tunnel
works and how the geotechnical control process protects public safety and adds value by ensuring an adequate
standard of design, site supervision and risk management is applied. The successful implementation of best
geotechnical risk management practice in the project is strongly influenced by the commitment of the project
client to follow the core element of the Joint Code of Practice for Risk Management of Tunnelling Works and to
follow up on the results of the independent auditing under the geotechnical control process.
1
1.1
INTRODUCTION
The details of the three contracts are described

below.
Project description
The Kowloon Canton Railway Corporations (KCRC)

Kowloon Southern Link (KSL) Project links West
Rails Nam Cheong Station and East Rails East Tsim
Sha Tsui (ETST) Station in the HKSAR. The 3.8 km
long railway will have one new station, West Kowloon
(WKN) Station.
The civil construction works of the project
were packaged into three Design and Build contracts. The total project cost, including railway systems, is expected to be about HK$8.3 billion. The
project is scheduled for completion in late 2009.
1.2 Contract KDB200

Contract KDB200 involves the section between ETST
Station and Jordan Road via the former Marine Police
Headquarters. WKN Station, two railway tunnels and
two emergency vertical shafts, namely Canton Road
Emergency Access Point (CREAP) and Peking Road
Emergency Egress Point (PREEP), were included
in this contract (see Figure 1 below). The contract
was awarded to Link 200 JV (formally known as the
Leighton Balfour Beatty Kumagai John Holland
Joint Venture).
Figure 1. Aerial photograph showing the alignment of Contract KDB200.
567
5000 horsepower. It was design in Germany and

its components were manufactured in Germany and
China and finally assembled in Guangzhou. The 50
steel cutting discs enable the TBM to cut through 1.5 m
of rock in about 40 minutes.
Notwithstanding the use of TBM, some works such
as ground investigation and grouting were required
under the contract on the already very busy road surface. Such works are to provide additional data for
geotechnical design and to mitigate the impacts before
construction of the relevant tunnel works sections.
The excavation for WKN Station was supported
by diaphragm walls, while that for the remaining cut
and cover tunnels was supported by temporary walls
comprising sheet piles, pipe piles and struts.
The most challenging part of these works from a
geotechnical perspective is the provision of a retrieval
shaft outside the Sheraton Hotel in Salisbury Road
serving as a works portal for the retrieval operation of the TBM. Traffic diversions were unavoidable.
This part of the works was constructed using the cutand-cover method. Temporary road decking has been
erected to facilitate the underground works and to
shorten the construction period.
1.3 Contracts KDB300 and KDB400
Figure 2. Slurry-type tunnel boring machine.
The geological sequence along the alignment generally comprises the following principal geotechnical
units:
fill (generally a few metres thick but up to 20 m
thick at the West Kowloon Reclamation site);
marine deposits and alluvium (described as
clayey/silty sand and sandy silt/clay with some
gravel, mainly found along Salisbury Road but also
present locally near Haiphong Road and undredged
pockets at the WKN Station);
residual soil and saprolite, mainly completely to
highly decomposed granite; and
moderately decomposed to fresh medium-grained
granite.
At the preliminary design stage, a cut-and-cover
option was planned for the tunnels. However, KCRC
decided to employ a Tunnel Boring Machine (TBM)
for the twin railway tunnels between the launching
shaft at the southern tip of WKN Station and the
retrieval shaft at Salisbury Road to minimize potential
disruption to road users, pedestrians, business operations and residents in the area. The twin railway
tunnels are approximately 8 m in diameter and 1.1 km
in length. The depth to tunnel crown varies from 8 m
to 24 m.
The slurry-type TBM (see Figure 2) uses 3800
kilowatts of electricity, which is equivalent to over
Contracts KDB300 and KDB400 involve the section between Jordan Road and Nam Cheong Station
of the KCRCs existing West Rail (WR). The two
contracts are divided at the Yau Ma Tei Ventilation
Building (YMTVB) (see Figure 3). The tunnel lengths
are 0.85 km and 1.06 km respectively. These contracts
were awarded to China State Construction Engineering
(Hong Kong) Limited.
In view of the ground conditions (mainly fill,
marine deposits, alluvium and completely decomposed granite), the depth of the tunnel (about 20 m
to soffit) and the lack of constraints on the ground
surface, the cut-and-cover tunnel method has been
adopted as the most suitable method in terms of
time and cost for the majority of the tunnel length.
The exception is the tunnel beneath Cherry Street,
which would be constructed using a mined tunnelling
method. The cut and cover excavation was supported
by temporary walls comprising sheet piles, pipe piles,
diaphragm walls, bored pile walls and struts.
The tunnel alignment is very close to many existing buildings, structures and utility services. Some
are very sensitive to construction-induced ground
movements such as the operating Mass Transit Railway tracks and buildings, highway bridge structures,
Drainage Services Department structures (e.g. box
culverts) and nearby buildings (e.g. the HSBC Centre). In order to monitor the effects of the construction
on the surrounding buildings, structures and utility services, extensive geotechnical instrumentation
568
Figure 3. Aerial photograph showing the alignment of Contracts KDB300 and KDB400.
has been installed, including ground/ structure/utility

settlement checkpoints, inclinometers, extensometers,
tiltmeters, vibration monitoring checkpoints, standpipes and piezometers. Instrument readings are monitored against alert, action and alarm values as defined
individually for particular instruments and sensitive
receivers.
2 THE ROLE OF THE GEOTECHNICAL
ENGINEERING OFFICE
2.1
Buildings ordinance and regulations
In the HKSAR, the regulatory control of building

works, in the interest of protecting public safety, is
by application of the Buildings Ordinance and Regulations (BOR). Under the BOR the definition of a
building includes, inter alia, any underground
space adapted or constructed for occupation or use
for any purpose including its associated access tunnels and shafts. Therefore, the development of private
underground space such as tunnels or caverns including their planning, design and construction all fall
under the BOR.
A KCRC project is considered a private project
in which the tunnel works may be exempted from
the administrative procedures of approval and consent
under the Buildings Ordinance. Under certain conditions, an Instrument of Exemption (IoE) would be
prepared and issued by the Building Authority (BA)
of the HKSAR Government. For the KSL Project,
because of the significant risk to life and property,
the design, risk management, construction and site
supervision would need to be implemented to a good
standard to protect public safety. Auditing of the
standard of geotechnical design, site supervision and
risk management is carried out by the Geotechnical
Engineering Office (GEO) of the Civil Engineering
and Development Department, as a technical adviser
to the Buildings Department (BD) of the HKSAR
Government.
2.2 Technical standards

GEO has issued a technical guidance note TGN24
(GEO 2005a), on specific aspects of site investigation for tunnel works in the HKSAR. It supplements
guidance on site investigation given in Geoguide 2
(GEO 1987) and Geoguide 4 (GEO 1992). The technical guidance note was prepared with the benefit of the
experience gained from the Harbour Area Treatment
Scheme Stage I (Pang et al 2006, Massey et al 2007).
GEO has also issued TGN25 (GEO 2005b), on the
implementation of geotechnical risk management in
relation to tunnel works. Tunnel works are defined as
tunnels, shafts, caverns and associated underground
facilities, however constructed. Construction of tunnel works may involve use of drill and blast methods, tunnel-boring machines, cut and cover methods,
techniques that incorporate insitu ground treatment,
groundwater control, installation of temporary and
permanent supports, etc.
The relevant Government departments and the profession were consulted in the preparation of these
guidance notes.
3
3.1
GEOTECHNICAL CONTROL PROCESS

Instrument of exemption
Within GEO a review panel has been set up to agree on

the geotechnical auditing requirements to be included
into the IoE, examine the key geotechnical aspects of
major submissions related to tunnel construction for
the project, and to oversee the standard of auditing.
The input started right from the planning stage of the
project in which KCRC demonstrated to GEO that they
had adequately identified and assessed the geotechnical risks, and had taken suitable risk mitigation
and control actions to manage the risks. In addition,
KCRC committed to adopt the Joint Code of Practice
(ABI & BTS 2004). The risks identified for the TBM
tunnels related to plant procurement, manufacture of
569
segmental linings, delay in launch shaft availability,

break-in to retrieval shaft, TBM assembly/removal and
operation including encountering adverse ground conditions, unforeseen underground obstructions, causing
excessive settlements or even damage to nearby buildings/structures/utilities during TBM operation and
interventions, compressed air blow-out along areas of
low soil cover and crossing over the operating MTR
tunnels.
In early 2005, KCRCs consultants submitted preliminary scheme designs for auditing by BD/GEO,
including a Geotechnical Basis of Design Report,
Ground Movement Prediction Report, Geotechnical
Instrumentation Report and an Existing Buildings and
Structures Report, supported by a Geotechnical Data
Report.
The BA issued the IoE to the KSL construction
works on 30 July 2005 pursuant to the Kowloon Canton Railway Corporation Ordinance. The exemption is
confined to only those procedures and requirements
relating to approval of plans, consent to commencement and resumption of works and occupation of
buildings provided under the BOR, such that the BAs
duties and sanctioning powers to ensure standards of
health and safety are not undermined.
Guidelines which KCRC were required to follow for
making submissions to GEO for geotechnical works,
including tunnel works, under the KSL project were
agreed and included into the IoE and summarized in a
Management Plan (KCRC 2005).
Under the agreement, KCRC had submitted the
design statements and various method statements for
the TBM tunnels, the cut and cover tunnels, CREAP,
PREEP, the launching shaft and the retrieval shaft to
BD/GEO for consultation.
KCRC had also submitted the excavation and lateral support (ELS) plans for the TBM tunnels where
there is excavation, loading/unloading of the ground
or changes to the groundwater regime. The critical
parameters shown in the TBM ELS plans include the
operating slurry pressures, the planned interventions
(use of free air or compressed air depends on the anticipated ground conditions) and the air pressures where
compressed air interventions are planned.
3.2
of the construction works is at a standard not inferior

to that required under the BOR. KCRC also employed
a team of Resident Site Staff (RSS), led by a Construction Manager to act as the Engineers Representative,
to supervise the works and the implementation of the
Project Risk Management Plan (KCRC 2006). In addition, AP, RSE, RGE, RGBC and RSC have jointly
prepared Site Supervision Plans (SSP) in accordance
with the Codes of Practice for Site Supervision (BD
2005a, b).
The Contractors were also required to appoint
Independent Checking Engineers (ICE) to provide
certification of consultation documents and verify
geotechnical design submissions for both permanent
and temporary works prior to forwarding to KCRC for
review under the contracts.
KCRC ensure that submissions are made at all
stages to BD, and to GEO where geotechnical
aspects are involved, in a timely manner prior to
the commencement of elements of the construction
works through the AP/RSE/RGE. KCRC and the
AP/RSE/RGE are required to ensure that all comments
given by BD and other relevant parties, in connection with their submitted consultation documents, have
been resolved to the satisfaction of BD and the party
concerned prior to commencement of construction of
the relevant part of the works.
3.3 During construction
During the course of works, the AP/RSE/RGE are
required to keep on site copies of certified working
plans, inspection and test records and other relevant reports for regular audit inspections by BD
and GEO. When significant changes in design
or method of working are necessary, then KCRC
through the AP/RSE/RGE are required to report this
to BD and GEO and ensure that all comments given
by BD and GEO are resolved to the satisfaction of BD.
KCRC through the AP/RSE/RGE are required to
report the following to BD and GEO immediately
when the following circumstances arise:
Construction accidents causing nuisance to the
public.
Irregularities causing inconvenience to the public
and/or damage to nearby property.
Construction non-conformities.
Other requirements for KCRC under the IoE
Under the IoE, KCRC is required to appoint Authorized Persons (AP), Registered Structural Engineers
(RSE) and Registered Geotechnical Engineers (RGE)
to co-ordinate the works and to certify the plans and
documents as well as completion of the works; and
to appoint Registered General Building Contractors
(RGBC), and Registered Specialist Contractors (RSC)
in the case of specialized railway construction works,
to supervise and carry out the relevant works.
KCRC is also required to instigate an assurance system and control scheme to ensure that management
Since award of the first contract for the KSL Project

in August 2005, monthly Buildings Ordinance Management Committee (BOMC) Meetings have been
held by KCRC with representatives from KCRC Management, the AP/RSE/RGE and from Government
including BD and GEO. One of the items of discussion
is the scope, standard and timing of the geotechnical submissions from the RGE to GEO. Geotechnical
auditing of the KSL Project is currently in progress.
570
ACKNOWLEGEMENTS
This paper is published with the approval of the Head of
the Geotechnical Engineering Office and the Director
of Civil Engineering and Development.
The authors acknowledge with gratitude the information and materials provided by the Kowloon-Canton
Railway Corporation and Link 200 JV for production
of this paper.
REFERENCES
ABI/BTS 2004. A Joint Code of Practice for the Procurement, Design and Construction of Tunnels and Associated Underground Structures [In the United Kingdom].
London: The Association of British Insurers/The British
Tunnelling Society.
BD 2005a. Code of Practice for Site Supervision 2005. Hong
Kong: Buildings Department.
BD 2005b. Technical Memorandum for Supervision Plans
2005. Hong Kong: Buildings Department.
GEO 1987. Guide to Site Investigation (Geoguide 2). Hong

Kong: Geotechnical Engineering Office.
GEO 1992. Guide to Cavern Engineering (Geoguide 4).
Hong Kong: Geotechnical Engineering Office.
GEO 2005a. Site Investigation for Tunnel Works (TGN24).
Hong Kong: Geotechnical Engineering Office.
GEO 2005b. Geotechnical Risk Management for Tunnel
Works (TGN25). Hong Kong: Geotechnical Engineering
Office.
KCRC 2005. Kowloon Southern Link Management Plan for
Compliance under the Buildings Ordinance Revision D.
Hong Kong: Kowloon-Canton Railway Corporation.
KCRC 2006. Kowloon Southern Link Project Risk Management Plan. Hong Kong: Kowloon-Canton Railway
Corporation.
Massey, J.B., Pang, P.L.R., Lo, J.Y.C. & Salisbury, D. 2007.
Developments in Tunnel Engineering in Hong Kong.
Proceedings of the HKIE Geotechnical Division Annual
Seminar 2007. Hong Kong: 137155.
Pang, P.L.R., Woodrow, L.K.R. & Massey, J.B. 2006. Development of geotechnical control and risk management for
tunnel works in Hong Kong. Proceedings of the HKIE
Geotechnical Division 26th Annual Seminar. Hong Kong:
7588.
571
Research on structural status of operating tunnel of metro in Shanghai and

treatment ideas
J.P. Li, R.L. Wang & J.Y. Yan
Shanghai Metro Operation Co., Ltd., Shanghai, P.R.China
ABSTRACT: Recently, with network operating and servicing time going on, the safety of operating tunnel
in Shanghai is becoming a focus in the circle of civil engineering. It is well known that safety of operating
tunnel is greatly influenced by the structural status of the tunnel. This article firstly give an introduction of main
diseases of operating tunnel based on the large amount of information and date collected in the past decades
about Shanghai Metro. Then causes for the disease are analyzed and related suggestions to prevent the diseases
deteriorate are given. It is shown that the main problem of operating tunnel include leakage, crack, longitudinal
settlement and constringency.
INTRODUCTION
Recently, with network operating and servicing time

going on, the safety of operating tunnel in Shanghai
is becoming a focus in the circle of civil engineering. However, only decades after shield tunnel was
introduced to China, few successful experiences can
be used for reference from domestic and abroad,
especially for Shanghai metro where tunnel was constructed in thick soft clay soil. This article firstly gives
an introduction of main disease of operating tunnel
based on the large amount of information and data collected in the past decades about Shanghai Metro. Then
causes for the disease are analyzed and the related suggestions for prevention are given. It is expected that
this article can be useful for those who is interested in
the health of tunnel of Shanghai metro.
2
2.1
MAIN DISEASES
of Shanghai metro belongs to the second or the third

level.
After decades of operating, the function of waterproof of sealing rod has already reduced to a low grade
for the non-uniform settlement of tunnel caused by
adjacent construction, pumping of ground-water, etc.
For the special stress state, most segment rings exhibit
an ellipse shape with horizontal radius enlarged and
vertical radius reduced. Correspondingly, the compressive stress at the hance of tunnel segment ring
decrease near outer surface and increase near interior surface. Gaps at the joint of segment rings maybe
formed. Since the sealing rod was set at the outer side
of segment ring, so the capability of water-proof was
reduced at the hance of the tunnel, and this lead to that
most leakage take place at the side of tunnel. In addition, for the great difference of structural style among
the shield tunnel, the station and the by-pass, differential settlement occurs at these positions and this also
leads to serious seepage.
Leakage
Based on the large amount of information collected

recently about Shanghai tunnel of metro, more than
ten seepage spots can be found at each section between
adjacent stations. Seepage spots in several sections
could up to fifty places. Figure 1 indicates that leakage
mainly occurs at both sides of tunnel. joints of station
and tunnel and the by-pass between up and down line.
Minority leakage places were founded at the hole of
bolt and the hole of grouting. According to technical
specification for water-proof of shield-driven tunnel
(DBJ08-50-96, 1996), Leakage of the operating tunnel
2.2
Segment crack
Few cracks are formed in the segment itself, but the

phenomenon that joint filler has pull-out can be found
for non-uniform settlement of operating tunnel which
makes the joints uncoupled. Most cracks located at the
corner of segment or unfilled corner take place during
the construction phase, such as production, maintenance, handling and consolidation. Additional, cracks
can be found between segment and track bed at the
position where large differential settlement take place
and where turning radius is small.
573
Figure 1. Leakage of the segment ring.
2.3
2.5 Convergence
Longitudinal settlement
Large settlement take place to the tunnel after decades

of operation, and the longitudinal settlement represents with regional characteristic. Figure 2 indicates
that the longitudinal settlement curve of Line 1 of
Shanghai metro contains two huge settle pits. One of
them is about 1400 m long, located at the Hengshan
road station and the maximum settlement is about
20 cm. The other one is located at the interval from
the South Huangpi road station to Shanghai Railway
station with the maximum settlement is nearly 30 cm.
2.4
Horizontal diameter of most segment ring enlarge from

2 cm to 4 cm, few of them even up to 7 cm, This
already exceed the design safety limit (1 5D).
According to statistics, segment ring with horizontal convergence deformation greater than 3 cm occupy
69.98% and those of greater than 6 cm occupy 6.63%.
The maximum convergence is about 15 cm and the gap
along the longitudinal joint is 11 mm which means that
circumferential bolt have already reach yield limit.
3
Rate of settlement
Figure 3 is duration curves of settlement of tunnel

near the People Square. Although the tunnel have been
constructed for more than ten years and the rate of settlement has being slowed down, it can not leads to that
the tunnel has already been stabilized for the creep
properties of Shanghai soft clay. How to control the
settlement is still a difficult problem faced by civil
engineering.
GENETIC ANALYSIS OF DISEASE
According to decades subway monitoring results, the

main factors related to the deformation of metro
structure are listed as following:
(1) Local Ground Subsidence
Based on the ground subsidence database of Shanghai city, it shows that the settlements of subway station
and tunnels are large if they located at the center
574
Figure 2. Longitudinal Settlement Curve of Line 1 of Shanghai Metro (1995.52007.12).
Figure 3. Duration curve of tunnel adjacent to People Square.
of subsidence region. Comparison of time-settlement

curves between ground surface and tunnel structure
measurement in People Square region gives great
agreement. Figure 4 shows the time-settlement curves
from 1999 to 2008. But the exacted ratio of tunnel
settlement to ground subsidence is not able to be
obtained.
operating safety standard eventually. There are some

typical cases presenting this phenomena, such as the
tunnel deformation near to West Ninghai Rd. pumping station of metro Line.1, leakage of water and sand
in pumping station of metro Line.2 crossing rive part,
Shilong Rd. station of metro Line 3, have been found
and treated against to the excessive settlement.
(2) Geology Conditions
(4) Maintenance and Operation Work
In Shanghai area, the soft soil stratum is about 30 m

deep from ground surface. The subway station and tunnel mainly located in soft soil. The soil is basically
saturated flowing-plastic or soft-plastic clay with low
shear strength (0.005 0.01 MPa), high water content
(above 40%), high compressibility (0.5 1.0 MPa1), sensitivity varying from 4 to 5, and rheological
behavior. In this very soft ground, the influence of
excavation and tunnel drive to environment could
not be ignored either for construction period or for
long-term operating period.
Vibrations of running trains would lead to tunnel settlement, and then water leakage will takes place due
to great tunnel settlement. In case the water leakage
becomes serious, it will induce more water/soil loss
and behave as larger settlement. The vicious circle is
formed in this way. The in-time and frequent maintenance work is a good way to prevent this vicious
circle. The research on long-term soil mechanics under
vibration of high consistercy and low frequency is still
undergoing. But it could be ensured that disadvantage
of the complicated soil behavior is the main factor of
tunnel operation safe.
(3) Quality of Construction

It would lead to large deformation if there were some
accidents occurrring during the construction of station and tunnel driving. The differential settlement
would develop along the operation life and overstep the
575
(5) Loading and Unloading activities near to tunnel

Due to the shortage of landing resource in urban environment like Shanghai city, the projects of foundation
pit located in the area of subway protection area tend to
Figure 4. Time-settlement curve of ground surface and tunnel.
be deeper, bigger, closer, more difficult and more risk.

As a consequence, it is necessary to pay more attention
to the influence of deep foundation pit excavation and
high-rise building to the subway station and tunnel.
area. Up to now, 30000 settlement monitoring points

and 1 convergence monitoring point every 5 segment
rings were set for metro Line1 to Line 4 which are all
operating metro lines in Shanghai.
(6) Groundwater exploitation, dewatering and pore

pressure reduction
(3) Disease Record Card System for Subway Structure
Soil consolidation induced by the groundwater

exploitation would also influence the subway structures.
Based on decades observation of longitudinal settlement curves of subway tunnel, positions with large
longitudinal settlement curvature are found. Disease
record card of subway structure were established to
make sure that inspection and monitoring could focus
on these positions. Then quick response and actions
could be taken according to the results of inspection
and monitoring in case of any abnormal situation.
CONTROL MEASURES
The safety guarantee system for subway structure

should take many aspects into account, including
the regulation, standard, monitoring management and
pre-alarm management.
(1) Regulation and Standard for Subway Control and
Protection
Supervised regulation should be established for subway line surveys, inspection and projects construction.
Process management and responsibility definition
should be enhanced to make sure that the whole
subway lines are under control.
(2) Standard of the Monitoring Procedure
It is the responsibility of monitoring engineer to monitor, inspect and analyze subway structure, which aims
to find the problem of structure in time and to guarantee the safety of subway. There are at least twice
settlement measurements and one convergence measurement for operating line; 3 4 times settlement
measurements and once convergence measurement for
new line; several times inspection for key positions.
More attention should be paid to the monitoring and
inspection of projects located in the subway protection
(4) Digital Information System

An overall digital scan for running subway and a big
GIS system are planning and constructing respectively.
The GIS system would control the safety of subway
structure and operation risk using information technology including inspects of geological conditions,
tunnel structure, waterproof system, settlement and
convergence.
(5) New technology and Equipment
The monitoring system with long distance, large
range, high precision and automatization characteristic
should be developed continuously. Remote monitoring
system and equipment configuration would be set on
the key position. The data, graphs and information are
able to be transferred and analyzed instantly.
(6) Project located in the Subway Protection Area
Seting protection area aims to reduce the influence
of loading and unloading effect from building and
excavation to tunnel by enhancing control standard
576
to the projects located in the subway protection

area.
(7) Scientific Research Project located in the Subway
Protection Area
The investigation to structural problem of metro,
including durability of tunnel, seepage and leakage,
settlement and convergence, has been preformed cooperating with research institute. Evaluation system and
index will be established to protect subway structure.
REFERENCES
Duan J.L., Tan Y.L., Wang Y., et al. 2006. Case study of foundation pits crossing subway tunnel. Chinese Journal of
Geotechnical Engineering 28(Z):18771879.
Li X.Z., Huang M.S. & Wang L.M. 2007. Experimental study
on creep and cyclic creep characteristics of saturated soft
clay. Journal of Chongqing Jianzhu University 29(2):56
59.
YeY.D., Zhu H.H. & Wang R.L. 2007. Analysis on the current
status of metro operating tunnel damage in soft ground
and its causes. Chinese Journal of Underground Space
and Engineering 3(1):157161.
Zhu Z.F., Tao X.M. & Xie H.S. 2006. The influence and
control of deep excavation on deformation of operating
metro tunnel. Chinese Journal of Underground Space and
Cui Z.D., Tang Y.Q., Lu C., et al. 2007. Prediction of ground

settlement induced due to changes in engineering environment in Shanghai. Journal of Engineering Geology
15(2):233236.
577
Maximising the potential of strain gauges: A Singapore perspective

N.H. Osborne, C.C. Ng & D.C. Chen
Deputy Project Manager, Land Transport Authority, Singapore
G.H. Tan
Managing Director, SysEng Pte Ltd, Singapore
J. Rudi
Instrumentation Engineer, Kiso Jiban Consultants Co. Ltd, Singapore Branch
K.M. Latt
Instrumentation Engineer, Soil & Foundation Pte Ltd, Singapore
ABSTRACT: Load monitoring of support struts in temporary excavations plays a crucial role in confirming
the stability and safety of the excavation. Data gained provides valuable feedback for the design engineer
to facilitate refinement of future designs. Much of this monitoring is undertaken by strain gauges, sensitive
instruments attached to the temporary supports, which are then linked to automated alarms through real time
systems. The success of the monitoring is directly linked to the performance and interpretation of the data
derived from the strain gauges, and the reliability of the real time system. There are a number of factors which
can interfere with the performance of the monitoring system, to successfully and usefully interpret the data; the
influence of these factors needs to be understood. The emphasis must be on the production of high quality data
that can be reliably processed and rapidly given to the end-user, such that erroneous readings are minimized
and genuine load changes are identified for interpretation. Through a number of case studies of deep excavation
projects in Singapore, influences on the monitoring system are reviewed, their potential impacts discussed, and
recommendations given to produce a high quality and reliable monitoring system, thus maximizing the potential
of strain gauges to be used for monitoring of performance.
1
1.1
INTRODUCTION
Importance of instrumentation
Instrumentation and monitoring has always played a

crucial role within the construction industry and recent
worldwide high profile construction failures have further raised its profile and importance. This rings very
true for Singapore. Technological advancements have
allowed the instrumentation industry to become more
sophisticated in how data is monitored, collected and
presented to the end-users. However the emphasis
remains on the production of quality data for useful
interpretation. Strain gauges are one very important
component of a fully integrated and comprehensive
monitoring system used to control the movements
and loads generated during excavations. In Singapore
they are used extensively for strut monitoring in deep
excavations and provide invaluable data ensuring that
construction control is maintained during excavation.
The build-up of load in the strut is monitored in real
time during construction and compared with the design

predictions at the various excavation stages. If significant discrepancies are observed, reanalysis is required,
and a review of the design assumptions undertaken.
However strain gauges are notoriously sensitive instruments and their readings can be influenced by a
number of factors, leading to misinterpretation when
the data is reviewed.
There are a number of wide-ranging potential influences impacting the accuracy of strain gauge results,
encompassing a number of related construction areas:
These include installation, environmental effects, construction activities and their position relative to structural members, all of which can result in erratic
changes in the strain gauge readings, some of which are
genuine load changes, some not. With the development
of real time systems, the results can be automatically transferred to the internet for interpretation or
sent to a mobile phone, but without any filtering of
the erroneous readings. If received by inexperienced
579
1.3 Choice of strain gauge type
Figure 1. Comparison of strain gauge and load cell results.
personnel, commonly, all readings are treated as genuine, with the potential to cause unnecessary false
panic and evacuation of the excavation, or erroneous
and ignoring genuine changes in load. Either scenario
results in a loss of confidence in the monitoring system. It is important therefore, that these very sensitive
instruments and their performance be fully understood during interpretation and every effort taken to
maximize the quality of the data.
1.2
Reliability of strain gauge a comparison with

load cell
Generally, load cell is known to provide reliable load

measurement as it measures the full load across the full
section of the strut. Strain gauge, on the other hand,
measure less then 1% of the strut cross sectional area
and then attribute that load to the whole strut. Figure 1
shows the data that gives an indication of the relative
accuracies of strain gauges and load cells. The graph
shows a general variance between 5% and 10%. Much
of this difference can be accounted for by their different responses to the various construction influences,
external factors such as temperature and EMI, and the
different ways they monitor load. A detailed discussion of this topic is beyond the scope of this paper and
needs to be part of a separate study. The correlation
between the two instruments, however, is sufficient to
indicate that strain gauge, is capable of monitoring the
struts accurately.
With this fact established, strain gauge has other
advantages over load cell. One fundamental benefit
is the cost effectiveness of strain gauge in Singapore;
with a strain gauge costing approximately U.S. $200
and a load cell considerably more at U.S. $4000. With
cost, normally to the industrys detriment, significantly
influencing the selection process, strain gauges are
the instrument of choice. Cost aside from a technical perspective, introduction of load cell in the strut
can create non typical loading conditions, but strain
gauge does not change the loading condition on the
strut. In addition, strain gauge can also be replaced
easily if damaged.
Once a strain gauge is chosen to monitor a strut,

there are two commonly available types for strut monitoring; spot weldable strain gauges (SWSG) or arc
weldable (AWSG), also referred to as surface mount
strain gauges. Both strain gauges are widely used in
Singapore, with slightly different attributes and advantages. SWSG have the advantage of being attached low
to the strut, therefore minimizing the impact caused
by bending error. However the AWSG is more robust,
its greater area fixed to the strut resulting in a better
area to area connection, making it in theory less sensitive to fluctuation from vibration caused by accidental
impact to the strut. By far the most important factor in
ensuring the performance of the strain gauges is that
they are installed correctly and not damaged during
installation.
For SWSG the gauge flanges are spot welded to
the strut, this requires exposing the thin 1.45mm
gauge, increasing the possibility of damage. After the
welding, the vibrating wire plucking coil housing is
mounted onto the gauge for measurement and as a form
of protective cover. It is crucial that the coil does not
touch the gauge, which can occur with some brands,
otherwise subsequent readings are affected. AWSG are
arc welded to the flange, however a dummy gauge
should always be used during welding to avoid damage, then replaced with the actual gauge, followed by
mounting the vibrating wire plucking coil and further
protection as required.The only potential problem with
this installation is the introduction of residual stress
into the structural member by the welding process. On
overall comparison of the respective merits of the two
gauge types the authors believe AWSG will provide
more reliable results a view shared by Broone &
Crawford (2000).
2
FACTORS INFLUENCING THE

PERFORMANCE OF STRAIN GAUGE
2.1 Location of strain gauge

When bending within the prop is likely to be significant, strain gauge reading will be affected by its
location in the prop. Strain gauges placed at the same
section may give different reading as they experience different stresses in the prop subject to bending.
Connections between the prop and waling result in
non-uniform stresses, as do connections to kingposts,
cross-bracing or runner beams. Thus strain gauges
installed near to these locations are likely to be
affected by the non-uniform stresses and will not give
a representation of the loads in the props.
2.2 Impact of electromagnetic interference
As vibrating wire strain gauges operate at a frequency between 600 to 1500 Hz, they are subject
580
Figure 2. Impact of EMI noise on strain gauge readings.
to electromagnetic interference (EMI). This compromises the accuracy of the readings by introducing noise
into the raw data, which can be very difficult to separate from genuine data, and therefore can be processed
and calculated as load. There are numerous potential
sources of EMI noise on construction sites including:
arc welding, machinery ignition, power generators and
power cables on the site. The noise takes one of two
forms, either as a general underlying trend impacting
the overall accuracy of data by increasing its spread. Or
as a high voltage surge, causing a spike in the load readings when, for example, a machine ignition is started,
electronic noise tends to lower readings, whereas magnetic noise increases them. With the advent of real time
monitoring and data processing at 10 minute intervals,
the impact of this interference becomes more significant. The first case leads to a general questioning the
accuracy of the strain gauge readings as the accuracy
range appears wider. The data spikes can result in an
alarm being breached, with the potential for work to
stop unnecessarily.
The impact of EMI noise can be clearly seen in Figure 2. During the working day on 9th & 10th September, electronic noise interference from a generator and
power cable caused a 200 kN fluctuating reduction in
load. The lunch hour can also be clearly seen when the
generator was turned off. By the 11th September the
noise had been identified and the generator removed,
hence more stable readings.
2.3 Temperature effect
The impact of temperature on the performance of strain
gauges installed on struts has long been recognized as
a significant factor affecting their performance, with
papers first published on the issue in the early 1960s
in Norway and Japan (NGI 1962 & Endo & Kawasaki
1963). Despite the advent of thermally matched strain
gauges, thermal influence persists. In the United Kingdom for an un-decked excavation subject to the thermal
effects due to sunlight, an annual temperature range of
50 C was measured leading to a variation of 2750 kN
in strut loads, after the base of excavation was reached
(Batten M et al. 1996). This equates to a change of load

of 65 kN per 1 C. Similarly in the United States significant variations were measured and reported. Boone
and Crawford 2000, recorded an 18.75 kN change per
1 C, with an annual temperature variation of 45 C on
their site resulting in an annual fluctuation of 844 kN.
The difference in these variations is due to the strut
area, in combination with the stiffness of the retaining
system, the ground it supports and the end restraint,
with a greater stiffness resulting in a greater impact
due to temperature variation.
Singapore lying 1.5 North of the equator experiences minimal seasonal variation in temperature, but
a significant diurnal range, with temperatures fluctuating from a low of 20 C to a high of 36 C. This poses
a different set of problems. The problem of temperature variation was first published by Niu et al (2005)
discussing fluctuations due to a deep excavation in
Singapore for the North East Line metro, in 1999.
A load change of 37 kN per 1 C was recorded, against
a theoretical change of 48 kN per 1 C. This being
calculated by:
Where P is change in load due to temperature

change (T), As is cross sectional area of the strut,
Es is Youngs modulus of steel and is the thermal
coefficient of expansion for steel.
For this excavation, 23% of the theoretical increase
in load was not observed in the instrumentation. The
absence of this monitored load was explained and
demonstrated by an outward movement of 2 mm of
the retaining system during the higher temperatures.
This phenomenon has been observed in the U.S. where
a potential 13 mm movement into the ground was
recorded in glacial till, (D. Druss 2000) and on other
projects in Singapore, with a 2 mm movement into the
ground of a 1.5 m thick diaphragm wall in soft Marine
Clay.
The same temperature phenomenon is seen consistently across deep excavations in Singapore. For a 25 m
deep excavation in soft Marine Clay on the Circle Line
project an increase of 30 kN per 1 C was measured,
Figure 3, across three different struts, over four days
of non excavation. This equates to only 56% of the theoretical increase being transferred to the strut load. In
this case the critical factor in mobilizing the full effect
of the temperature lies in the ground.The retaining system was very stiff, with 1.5 m thick diaphragm walls,
compared to the NELP example where a soldier pile
system in considerably stiffer ground was used. Further to the general trend in the figure, there is some
scatter in the data, this is due to readings taken at
between 8:00 and 9:00 am and attributed to the very
localized effect of plant start-up up causing EMI.
581
Figure 3. Impact of temperature on strut load.
Figure 5. Impact of welding on adjacent strain gauges on

the same strut.
by worker, which can either permanently damage the

gauge or partially damage the gauge or connection.
2.6 Welding
Figure 4. Loss of preload after strut installation.
2.4
Preloading
Preloading has an important bearing on the strain

gauge performance in early stages. The fundamental
action is to install the strain gauge, take base readings
and check it is working correctly prior to the onset
of preloading. Base readings should ideally be taken
prior to the strut being connected to its end supports.
During pre loading the load registered in the strain
gauge and load cell should be checked against the jack
load; however loads cannot be expected to match perfectly. The jack registers the highest load, followed by
the strain gauge and then the load cell; this sequence
tends to be the case observed in Singapore.
This can be attributed to a number of factors: the relative positions of the instrument, relatively low loads
being used; construction difficulties in placing the
strut truly perpendicular to the retaining system; temperature effects; the introduction of load cell in the
struts affecting the overall load transfer pattern.
2.5
Disturbance by construction activities
As with all excavations, the accuracy of the instrumentation readings can be accidentally influenced by
a number of different construction activities, leading
to potential misinterpretation of the readings. One of
the obvious is the accidental damage of strain gauge
One of the construction effects resulting in erratic

and significantly high loads registered by the strain
gauge is welding. High heat from welding of horizontal ties or lacings to the strut member, which usually
commences after preloading, can result in a high and
sudden increase in the strut loads. As shown in Figure 5, the welding of lacing to a strut has caused the
strut load measured by a pair of strain gauges located
close to the lacing, to rise suddenly.
The impact of this welding on the strut load is clearly
evident.
As shown in Figure 5, the readings on both sides
of the strut web show a sudden drop, probably associated with EMI noise, followed by a sharp increase of
load on the welding side. An increase from 1300 kN
to 1900 kN was recorded. On the non-welding side a
minimal rise in load was registered. On completion
of welding the impacted gauge did not recover to its
original load but remained at its elevated level, which
is not representative of the overall load in the strut.
This residual stress, recognized since 1964, is not representative of the actual load of the strut, and if clearly
identified from the readings and construction activity,
the reading can be adjusted to account for this effect.
2.7 Casting of permanent slab
Another construction impact on temporary supports,
and their strain gauges, is the effect due to casting
of permanent components of a top-down excavation.
During the casting of a 1.5 m thick roof slab, the impact
of the curing and expansion of the slab can clearly be
seen in the two layers of struts above the roof slab,
Figure 6. A significant drop of 500 kN was observed
across the full excavation, followed by an increase several days later and a return to the ongoing trend of the
582
The simile, a chain is only as strong as its weakest

link, rings very true when applied to any real time system. Any failure of any component within the system
compromises the whole monitoring scheme leading
to an absence of results to the end user. Apart from
the strain gauges, the potential numbers of points
that can fail within the system are numerous. These
include the cabling, the data logger itself, the phone
system, the power and the server. Failures of all of
these components have been experienced.
3
Figure 6. Impact of casting of permanent roof slab on strut
loads above.
3.1
load. Once clearly identified, this phenomenon can be
easily linked to construction activity, and not used to
cast doubt on the accuracy of the strain gauge results.
2.8
Negative values
Another phenomenon seen in strain gauge in deep

excavation, particularly in soft clay, is that of negative loads in the top strut layers, indicating that the
struts are in tension. This is frequently blamed on the
instruments themselves and regarded as error readings.
However investigations into a number of these cases
have identified that the strain gauges are functioning
well and recording genuine loads. Independent checks
through cut off tests and inserting jacks have demonstrated that these struts are in tension. This is due to a
combination of factors: the loads in the struts tend to be
originally low at higher levels; loss of preload; and the
deflection profile of the retaining wall. With soft clays,
significant retaining wall movements are recorded,
with deep seated movements occurring below excavation levels. As the excavation progresses, stiffer struts
with greater preloads are used. Combined with these
large movements below the struts, the retaining wall
can rotate about the strut, resulting in a small backward movement into the soil at higher level. This is
also reflected in the inclinometer readings.
2.9
Problem related to real time system
To make the most effective use of strain gauge data, for

deep excavations, it is prudent to link the instruments
via data logger to the office computer and mobile
phones in a real time system. However by implementing such a system two noteworthy problems need to be
considered.
First the potential high number of alert alarms generated. The erroneous alerts can lead to a loss of
confidence in the system and potentially a genuine
alert being lost amongst the false alarms.
The second and potentially more serious problem
lies with the robustness of the real time system itself.
HOW TO MAXIMIZE THE POTENTIAL OF

STRAIN GAUGE
Right location of strain gauge on strut
There are two important considerations when locating a strain gauge on a strut. First its location on the
strut, and then its position relative to other structural
members. If bending within the prop is likely to be
significant, strain gauges should be located to account
for it, four gauges for a circular prop or proportionately spaced along the web for an I-beam. Connections
between the prop and waling result in non-uniform
stresses, as do connections to kingposts, cross-bracing
or runner beams. Therefore the gauge locations should
be at maximum distance from these areas, to be fully
representative of the loads passing through the strut.
3.2
Cross referencing with load cell
British CIRIA C517s guide on temporary design

(Twine & Roscoe 1999) recommends strain gauges
over load cells. However as cross referencing of data
from different instrument types is critical to gaining a
full picture of the excavation induced movement, it is
recommended that at least some load cells be included
in the instrumentation design. This concept is fully
recognized in Singapore.
The Building Control Authority of Singapore
(BCA) states that deep excavations must be monitored
by a combination of strain gauges and loads cells.
Land Transport Authority of Singapore (LTA), the
client for the majority of deep excavations in Singapore
which are for the continually growing underground
rail network, specifies that as a minimum requirement, 25% of all struts for deep excavations shall be
monitored for load by strain gauge and/or load cell, and
of those monitored by strain gauge, 15% shall also be
monitored by load cell.
3.3
Minimizing EMI effect
EMI noise can be avoided by a few simple measures

on site. It is recommended that cabling lengths be kept
to a minimum; the cables and datalogger be located
at least 5 m away from any potential source of EMI.
Figure 7 shows a typical layout of data logger in an
excavation site. Regular checks should be undertaken
583
Figure 7. Typical layout of data logger on an excavation site.
by a portable EMI meter to ensure that cable and datalogger zones remain EMI free. Although cables can
be protected from electronic noise, cable joints and
the inevitable on site cutting and re-splicing of cables
are weak points that can be easily corrupted, therefore
particular attention should be paid to these locations.
strongly recommended that the capacity for data transfer of any such system is in minutes and that a wireless
system be utilized.
An understanding of the potential problems can
reduce false alerts, combined with alerts going only
to knowledgeable personnel who are fully cognizant
of the construction works being undertaken.
Potential failure of the cabling, the data logger, the
phone system, the power and the server within the system can lead to loss of data or delay in transmission of
data. To ensure that the system is fully automated and
seamless, all these areas need to be rigorously checked
and fail safes written into the systems to inform the
system manager if any of these components fail, rather
than assuming that all are functioning smoothly.
3.4 Account for temperature effect
The temperature range is less in Singapore, and therefore the impact of load less than elsewhere in the
world. However strut loads are very closely linked
with monitoring control of excavations. The maximum
design load of a strut, using moderately conservative
soil parameters, is used as the work suspension level,
with an alert set at 70% of this capacity. There can be
significant implications on the work, including stoppage of works if the temperature effect is not properly
accounted for in the strut loads. A number of different
solutions include painting struts white and daily spraying to reduce temperature impact. It is suggested that
the most appropriate solution is to account for the theoretical temperature effects during design and to add
them to the monitoring control values to ensure that
work is not impacted unnecessarily.
It is clearly evident that strain gauges are essential in

monitoring and controlling deep excavations. With the
increasing sophistication of the real time systems producing vast quantities of data, combined with M2M
capabilities allowing automated alerts, and strict alarm
limits on the monitored loads within the temporary
structural systems, the results from strain gauges are
under very close scrutiny. Therefore quality data and
a clear understanding of both strain gauges and how
the construction activities impact that data are crucial
to the interpretation of strain gauge results. Without
this, confidence in the performance of the system
is lost, resulting in results being ignored as errors,
very dangerous in the monitoring environment, or
numerous unnecessary alarms received impacting the
construction progress.
To maximize the potential of strain gauges, their
locations and that of the cabling must be planned,
installation carried out by skilled personnel aware of
the problems and the potential to compromise data.
Data interpretation should also be by skilled personnel
fully aware of the design predictions for the excavation, the excavation progress and the potential impact
of the excavation on the strain gauge results. Finally the
processing system taking data from the strain gauge to
the end user must be seamless and robust such that
this component does not fail leading to a complete
breakdown of the whole system.
3.5
Protection against disturbance by construction

activities
Damage to strain gauge and disturbance on the data

due to construction activities can be avoided through a
combination of adequate protection, education of site
staff through tools box talks and clear well marked
signs on site. If damage occurs on a frequent basis,
the more extreme measure of fining individuals or
companies can be considered.
3.6
Maximizes strain gauge reading with real

time system
To maximize strain gauge reading for monitoring of

deep excavations, it is prudent to link the instruments
via data logger to the office computer and mobile
phones in a seamless fully automated machine to
machine (M2M) system where no human intervention
is required for onward transmission of results, resulting in a fully real time system (Tan et al 2004). It is
CONCLUSIONS
REFERENCES
Batten, M., Powrie, W., Boorman, R. & Yu, H. 1996. Measurement of Prop Loads in a large excavation during construction of the JLE station at Canada Water, East London.
Proceedings of Geotechnical Aspects of Underground
Construction in Soft Ground, London 1996 pp 712
584
Boone, S.J. & Crawford, A.M. 2000. The effects of temperature and Use of Vibrating Wire Strain Gauges for Braced
Excavations.Geotechnical News September 2000.
Druss, D.L. 2000. Discussion The effects of temperature
and Use of Vibrating Wire Strain Gauges for Braced
Excavations, Geotechnical News Vol. 18 No. 4 pp 24.
Endo, M. & Kawasaki, T. 1963. Study of Thermal Stresses
Acting on Struts. Transactions of the Architectural Institute of Japan No. 63 pp 689692
NGI. 1962. Vibrating Wire measuring Devices Used at Strutted Excavations, Technical Report No. 9. Norwegian
Geotechnical Institute, Oslo.
Niu, J.X., Wong, I.H. & Makino, M. 2005. Temperature

Effects on Strut Loads and Ground Movements for a
31m deep excavation in Singapore. Proceedings of Underground Singapore 2005.
Tan, G.H., Ng, T.G. & Brownjohn, J. 2004. Real Time
monitoring and Alert Systems for Civil engineering Applications using Machine to machine Technologies. Int. Conference on Structural and Foundation Failures, Singapore
2004.
Twine, D. & Roscoe, H. 1999. Temporary Propping of Deep
Excavations, CIRIA Guide C517.
585
Discussion on design method for retaining structures of metro

station deep excavations in Shanghai
R. Wang, G.B. Liu & D.P. Liu
Department of Geotechnical Engineering of Tongji University, Shanghai, P.R. China
Z.Z. Ma
Shanghai Shentong Holding Co., Ltd, Shanghai, P.R. China
ABSTRACT: By investigating many projects in Shanghai metro construction, notable differences are found
to exist in present design schemes, especially in steel bar content of diaphragm wall and struts load. In this
paper, some design schemes of retaining structures for metro station deep excavations which have the similar
geological and structural situations are compared. Their rationality is evaluated and tested according to the
results of field monitoring (e.g. strut load, bending moment of diaphragm wall deduced from lateral deformation
curves). By adopting these measurements, a series of software, calculating models, methods and parameters can
be summarized. Some key factors which affect the correctness of the design are emphasized. Finally, suggestions
that satisfy demands of safety and economy are given. These suggestions help improve the design of retaining
structures for deep excavation so that the design standard in Shanghai can be unified gradually.
In order to prepare the 2010 EXPO, construction

activities of civil infrastructures and buildings in the
municipality of Shanghai have been increased significantly. To meet the demands of high efficiency city
traffic, extensive underground metro system in Shanghai has been rapidly constructed. How to accomplish
the target of completing 400 km network in both quick
and satisfactory way under the situation of large scale
and high risk becomes a question to the constructors
in Shanghai. Many problems during construction have
been researched by many researchers and engineers
over the years (Liu & Hou 1997; Liu et al. 1999; Liu
et al. 2000, 2001). But few reports about the design of
retaining structures of excavations are analyzed. Some
retaining structure design schemes of normal metro
station excavations which have the similar geological and structural situations are selected. According to
the results of field monitoring (e.g. strut load, bending
moment in diaphragm wall deduced from lateral deformation curves), the rationality of designs are analyzed.
A series of software, calculating models, methods and
parameters can be summarized, and key factors affecting the correctness of the design are emphasized. The
intention of this paper is optimizing design methods

satisfy demands of safety and economy. These suggestions help improve design of retaining structures
for deep excavations so that the design standard in
Shanghai can be unified gradually.
2
2.1
DESIGN OF RETAINING STRUCTURES

OF DEEP EXCAVATION
Background of the design
2.1.1 Shape of the excavation

Compared with the diversification of common excavations, almost all metro station deep excavations are
similar. The main excavation of metro station looks
like a dumbbell on the plane, which divided into three
partsstandard segment in the middle and end wells
at two edges of the excavation.
2.1.2 Geological condition
Under the ground of Shanghai, there is soft soil about
dozens meter deep, whose water content, degree of
sensitivity, compressibility and rheology are high, and
the unit weight, strength and permeability are low.
587
Table 1. The control criterion of environment protection in excavation.

Protection
grade
The requirement of maximum ground settlement and

maximum retaining wall horizontal displacement
Maximum settlement of ground 0.1%H*;

Maximum horizontal displacement of retaining
wall 0.14%H;
Ks** 1.8
Maximum settlement of ground 0.2%H;
wall 0.3%H;
Ks 1.6
Maximum settlement of ground 0.5%H;
wall 0.7%H;
Ks 1.4
The requirement of surrounding protection

In the area not more than 0.7H away from excavation,
there are metro, municipal pipes, gas pipes,
important water pipes and so on. Those important
building and utilities must be secured.
In the area 12H away from excavation, there are
important mainlines, water pipes, important in-used
structures and buildings.
In the area not more than 2H away from excavation,
there are less important branch lines and general
buildings and installations.
* H is the depth of excavation; ** Ks is the safety coefficient against basal heave.
2.2
Design method of the retaining

structure in Shanghai
2.2.1 Excavation protective grades

According to the engineering experience of Shanghai
metro deep excavation engineering and the requirement of the surrounding environmental protection
(Shanghai standard foundation design code DGJ0811-1999), the deformation control criterion of excavation is classified into three protective grades (Table 1).
2.2.2 Model
Finite element method (FEM) of two-dimensional vertical elastic subgrade beam was adopted for analyzing
the deflection of diaphragm wall. In the FEM modeling, lateral earth pressure was taken as triangular
and rectangular. The distribution of horizontal load
is slope gradient on the excavation plane and rectangular under the excavation plane. The distribution
of horizontal spring stiffness coefficient of the passive zone soil was also considered as two distributions.
They have triangular distribution with slope gradient
within effective depth under the excavation level and
rectangular distribution below the depth (Figure 1).
Figure 1. Vertical elastic subgrade beam model.
underground stations as the objects of analysis, the

differences of them are as follows:
3.1 Steel bar content of diaphragm wall
DIVERSITY OF THE DESIGN RESULTS
Considering the similarity of excavation form and geological condition of metro stations in Shanghai, the
retaining structures design of metro station deep excavations should be unified. But it is to find notable
differences existing in the present design schemes by
investigating many projects in Shanghai metro construction, especially in steel bar content of diaphragm
wall and struts parameters. Results of retaining system
design of fifteen deep excavations of metro stations
in Shanghai are listed in Table 2. Regard two layers
The excavation depth is at the range of about 1517 m

in standard segment of two layers underground excavation, but the unit weight of longitudinal bars plus hoops
of diaphragm wall is quite distinction, from the minimum 127.05 kg/m3 to the maximum 232.63 kg/m3 , the
value discrepancy is about 83%.
Station 1 and 4 are similar in excavation depth,
but about 22% discrepancy in steel bar content of
diaphragm wall.
The struts number of cross section of station 8 is one
more than that of station 2, but still 44% higher steel
bar content of diaphragm wall than that of station 2.
588
Table 2.
Design results of bracing structural system of some stations of metro line M1.
Standard segment of the excavation
Serial
number of
station
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
Structure
form
Excavation
depth/m
Length of the
diaphragm
wall/m
Thickness of
the diaphragm
wall/m
Embedded
ratio of the
diaphragm wall
Steel bar content

of the diaphragm
wall/kg m3
Vertical
struts
number
Tow layers
underground
17.28
16.80
16.60
17.29
16.30
16.30
15.66
16.44
15.57
15.59
14.56
14.92
18.37
24.24
21.56
34.0
30.0
32.0
31.0
30.0
28.5
30.0
30.0
28.5
28.0
28.6
28.3
32.0
43.0
37.0
0.8
0.8
0.8
0.8
0.8
0.8
0.8
0.8
0.6
0.7
0.6
0.6
0.8
1.0
1.0
0.97
0.79
0.93
0.79
0.84
0.75
0.92
0.82
0.83
0.80
0.96
0.90
0.74
0.77
0.72
127.05
128.86
145.82
155.66
155.85
156.43
169.47
185.58
186.44
192.56
196.97
232.63
139.80
145.21
153.10
5
4
5
5
5
5
4
5
4
4
4
4
5
7
6
Three layers
underground
*Note: Steel bar content of the diaphragm wall in table 1 refers the unit weight of longitudinal bars plus hoops.
the volume when the excavation depth and length of

the diaphragm wall increase. So the steel bar content
of diaphragm wall decreased.
3.2 Design parameters of steel struts
3.2.1 Comparison of design parameters of different
stations
Strut design parameters of 3 stations which have the
closely excavation depth and stratum distribution are
compared (Table 3).
Spacing of the struts often ascertained according to
experience, so there are less differential in it. But the
design struts load calculating by model is discrepancy
obviously.
Figure 2. Variation of steel bar content of diaphragm wall

with excavation depth.
The thickness of diaphragm wall of station 10 is

10 cm thicker than that of station 9, but almost the
same in steel bar content.
All above shows that the regularity of steel bar
content of diaphragm wall is not obvious now.
In addition, the decreasing tendency of steel bar
content of diaphragm wall along with the excavation
depth increasing is shown in Figure 2.
The steel bar content of diaphragm wall equals to
Ag divided by V, which Ag is the total weight of longitudinal bars plus hoops of a diaphragm wall; V is the
volume of a diaphragm wall. According to the design
schemes, the diameter and spacing of the longitudinal
bars and hoops are similar. But the spacing of both bars
increases at the range of approaching wall toe; it causes
the increment of weight of steel bars less than that of
3.2.2 Comparison of struts load by designing and

by field monitoring
Field monitoring is an important means of providing
immediate feedback to designers during excavation
and of documenting the actual performance as a case
history for future references.
Testing by the monitoring data of station 9 and 13
in Table 2 after excavation, real struts load are far less
than the designing.
Station 9: The percentages occupied by real struts
load to design values are 67% (level 2); 58% (level 3)
and 67% (level 4).
Station 13: The percentages occupied by real struts
load to design values are 72% (level 1); 69% (level 2);
46% (level 3); 37% (level 4) and 36% (level 5).
Considerable bearing capacity is wasted especially
in the lower struts. So the struts design should be
optimized.
589
Table 3.
Comparison of struts parameters of three station excavations (Standard segment).
Station
II
III
Excavation depth/m
Spacing on plane/m
Vertical spacing/m
15.57
about 3.0
0.5
3.9
3.9
3.1
3.07
/
1500
2000
1300
14.92
about 3.0
0.4
4.34
3.7
3.42
2.6
800
1660
2300
1975
15.59
about 3.0
1.0
4.24
4.1
3.4
2.85
800
1850
2000
1350
Struts load of design/kN
Level 1ground level

Level 2Level 1
Level 3Level 2
Level 4Level 3
Final levelLevel 4
Level 1
Level 2
Level 3
Level 4
Figure 4. Usage percent of software in calculating the main

body excavations.
4.3 Relationship between design and construction

Figure 3. Struts load comparison of designing with field
monitoring.
FACTORS ANALYSIS
By investigating eight design institutes which take on

most design work of metro projects in Shanghai, there
are many factors causing the disperse of design results.
The main reasons are as follows:
4.1
Construction is the means to realize the design

schemes. It can test the rationality of the design.
Therefore, mutual cooperation can help both of them
getting better. But little contact between them makes
the design theoretical; say nothing of optimizing
the design according to the feedback of the field
construction.
5
Diversification of calculating tools
The software calculating the internal force and lateral

deflection of diaphragm wall is not unified. SUPERSAP, ANSYS and MIDAS are used in calculating
the main body of the excavation; FRWS, LIZHENG,
YAO1, YAO2 etc. are used in checking the results or
designing the affiliating structures.
4.2 Values of the parameters
Effected by the designers subjective understanding of
the code, parameters like overload, lateral pressure,
restrictions and so on are adopted differently. The difference of these initial parameters will lead to the
different calculating results directly.
In addition, the phenomena of the designers concentrating on safe while ignoring the economy made
the design schemes are quite conservative.
DESIGN SUGGESTION
According to the design code and practical experience of Shanghai, some key points about the design
of the retaining structure of normal metro station deep
excavation are suggested or emphasized:
5.1 Stricter design demands
The load coefficient has already increased from 1.25 to
1.35; The structure significance coefficient 1.1 should
be considered; The concrete protection layer is thicker
than before.
5.2 Value of kv
The restriction under the toe of the diaphragm wall
is simulated by vertical spring which kv often adopt
10000 kN/m3 in Shanghai.
590
Figure 5. Empirical value of k in Shanghai.
5.3
Calculation of lateral loads
Results of lateral loads dont have much difference no

matter the earth and water pressure calculated together
or not in Shanghai. But it had better to calculate them
separately.
5.4
Earth pressure
Earth pressure is a key parameter of excavation design.

In the practices of soft clays excavation engineering
in Shanghai, by means of a lot of field measurements,
indoors tests and centrifugal mode tests, the relation of
earth pressure with different factors such as retaining
walls displacement, time, and construction parameters is erected. A practical formula of calculating earth
pressure has already been put forward. The active earth
pressure coefficient k can be ascertained by Figure 5
after excavating to the final level.
5.5
Equivalent subgrade coefficient Kh
Kh is a main calculation parameter in the calculation

method of displacement and internal force of retaining wall used by design institutes. An equivalent level
resistance coefficient Kh is used to reflect the capacity that soil resists deformation. It is a coefficient
that involves viscidity, elasticity, plasticity and other
diversified construction factors.
For purpose of facilitating design and engineering application, according to practices in Shanghai,
usually the following simplified formula can be adopt:
when the subgrade without improvement:
Figure 6. Maximum struts load by field monitoring.
Where i , ci , i is specific gravity, cohesion, angle of

internal friction of layer i respectively; hi and hi1 are
bottom depth of layer i and layer i 1 respectively; Bj ,
Tj , Hj is excavation width, excavation time, excavation
depth of progress j respectively.
when the subgrade with improvement:
5.6 Struts load

By survey of 26 station excavations, the maximum
monitoring loads of each level struts are suggested.
According to Figure 6, most values of struts load of
excavation in Shanghai are 5001000 kN (level 1),
591
Figure 7. Deflection ratio of the wall in different cases and

wall thickness.
Figure 8. Comparison of deflection, bending moment in

different wall thickness (excavate to final level).
12001800 kN (level 2), 12002200 kN (level 3),

10002500 kN (level 4 and 5) respectively. The maximum values are about 1300 kN (level 1), 2800 kN
(level 2), 3000 kN (level 3), 3000 kN (level 4 and 5)
by deleting some special factors.
As shown in Figure 7, although the deflection of the
diaphragm wall gradually increasing along with the
excavation deeper, the increment of the deflection
become lower with the wall getting thicker. But the
method of confining the deflection by simply increasing the wall thickness is neither economical nor
effective. It is not only playing the invalid function to
limit the deflection but causing the bending moment
of the wall increasing obviously when the thickness
increase to specific value. (Figure 8) So the thinner
wall should be better when the demand of safety can
be satisfied. The thickness of the diaphragm wall in
Shanghai often adopted 600 mm, 800 mm, 1000 mm
and 1200 mm which increment is 200 mm. As the construction machine developed, thickness increment by
100 mm as modulus can be realized. Nowadays the
attempt to apply the 700 mm diaphragm walls has
succeeded in metro line m1 in Shanghai.
Figure 9. Video, data collection and net page platform of

the remote monitoring system.
5.7 Remote monitoring, management and

consulting system by use of internet technology
Remote monitoring and management system is intelligentized monitor, measurement and management
system based on network technology and it is built
on local network and Internet. Such a system can deal
with the disperse projects, make the information, manpower and other resource of relative companies shared
each other, make it convenient to follow the project and
know what is going on.
People taking on monitoring of the construction
field inputs the monitoring data into the system
592
immediately after monitoring. The project managers

and designers can consult the data through a specific
net page platform; if there is any risk existing in the
project, some construction methods or design adjustment can be done to prevent the accident happen. It
will bring utmost profit to companies by providing
intelligent and effective way of project management.
At the same time, it can be also convenient to build
and manage database and provide remote technology
consultation.
6
CONCLUSION
Based on the survey of large quantity of monitoring

data of the retaining structure of metro station deep
excavations in Shanghai, the following conclusions
can be drawn:
Differences in the design schemes of the retaining
structure of normal metro station deep excavation is
depicted and analyzed. It mainly focuses on the steel
bar content and steel strut load, which disaccordance
with the character of similarity and regularity of metro
station excavation.
A lot of factors affect the design results. Difference
of calculating software, initial inputting parameters,
designer objective understanding and so on becomes
the main reasons.
Suggestions are given according to the code and

practical experience to help unify the design.
REFERENCES
Liu, G.B., Huang, Y.X. & Hou, X.Y. 2001. Study on Equivalent Lateral Resistance Factor of Soils in Excavation,
China Civil Engineering Journal.
Liu, J.H., Liu, G.B. & Fan, Y.Q. 1999. Theory of Time and
Space Effect and its Practice in Soft Soil, Journal of
Underground Engineering and Tunneling.
Liu, J.H. & Hou, X.Y. 1997. Manual of Excavation engineering, Chinese Construction Industry Press, Beijing.
Series of New Technologies of Environment Protection in
the Construction of Subway; from Shanghai Subway
Corporation (General).
Shanghai Construction Commission. 1999. Shanghai standard foundation design code (DGJ08-11-1999), Shanghai
Construction Commission (SCC), Shanghai, China. [In
Chinese].
Sun, G.S. & Zheng, D.T. 1987. Soft foundations and underground engineering, China Architecture & Building Press,
Beijing, China.
593
Risk analysis for cutterhead failure of composite EPB shield based on

fuzzy fault tree
Y.R. Yan & H.W. Huang
Q.F. Hu
Shanghai Institute of Disaster Prevention and Relief, Shanghai, P.R. China
ABSTRACT: All factors and related hazards in the cutterhead of composite EPB shield were systematically
analyzed, and then fuzzy Fault Tree Analysis (FTA) model of cutterhead failure was proposed. Main basic events
affecting the occurrence probability of the top event were verified by a quantitative analysis, which could be
applied in the risk analysis of EPB shield machines cutterhead. Compared with the traditional fault tree analysis,
the fuzzy fault tree method can get the cutterhead failure possibility distribution of composite EPB shield. At
last, the measures that help reducing the cutterhead failure occurrence were presented.
1.1 The actuality of composite EPB shield

cutterhead failure
With the manufacture technique development of shield
tunnelling, the scope of strata in which shields works
becomes more and more widely. Shield technology
is not only used in relatively uniformity or single
ground, but also used in mixed ground that changes
from hard rock to mixed face and soft ground (and vice
versa) at the tunnel level. During the excavation, the
highly abrasive and frequently changing mixed face
ground causes high cutter wear, especially flat cutter wear; the accident rate of cutter disc to the total
shield is highly to one half. The main forms of accidents are cutter abrasion, cutter disc abrasion, cutter
disc distortion and so on (Lei Guo 2006). These accidents made the advance rate and cutterhead service life
largely depressed. According to statistics, in domestic
composite EPB (Earth Pressure Balance) shield construction, the failure of cutter abrasion came forth at
differently degree (Weibin Zhu et al. 2006).
event is known. Fault tree analysis is based on Boolean

algebra. A quantitative analysis needs probabilities of
all basic events or the minimal cut set, which are mostly
obtained by statistical data or subjective judgmental
data based on experts experiences. These data have
uncertainty because of various influence factors during
statistical procedures and limitation of experts experiences. It is necessary to define a fuzzy value in the
probabilistic space to represent a single probability.
Basic concepts and methods of the fuzzy fault tree
were proposed in 1980s (Tanaka et al.1983, Furuta
1984). At present, the study of the fuzzy fault tree is
almost focused on algorithm and the integrated theory systems have not been established and verified in
practice (Singer 1999). A fuzzy set is introduced in
this paper and the failure probability of basic events
is replaced by the fuzzy failure probability. A triangular fuzzy number is introduced to represent the failure
probability of a basic event and the fuzzy failure probability of a top event is obtained by fuzzy number
operation. Measures to reduce the cutterhead failure
possibility of the EPB shield are verified by analyzing
the importance of basic events.
1.2 The introduction of fuzzy FTA

In the traditional fault tree, a failure event system (top
event) is divided into many sub-events with a combination of series and parallel. Its failure probability can be
back-calculated according to the logical relationship of
the fault tree when the failure probability of each basic
2
2.1
FUZZY SET THEORY AND ITS OPERATIONS

Fuzzy set
Fuzzy set theory was introduced by Zadeh (1965) to

deal with the problem in which the phenomena are
595
imprecise and vague. Let X be a collection of objects,

called the universe, whose elements are denoted by x.
A fuzzy subset A inX is characterized by a membership
function fA (X )which associates with each element x
in X a real number in the interval [0, 1]. The function
value fA (X ) represents the grade of membership of x
in A. The larger the fA (X ) is, the stronger the degree of
belongingness for x in A.
2.2
Fuzzy numbers and its operations
Switch between the fuzzy numbers are used to handle

imprecise information such as close to 5, high reliability, low failure rate, etc. There are many forms
of fuzzy numbers to represent the linguistic values.
In here, triangular fuzzy numbers are applied. Let x,
a, m, b R (real line). A triangular fuzzy number is a
fuzzy number A in R, if its membership function fA :
R [0, 1] is
In this study, triangular fuzzy numbers are

employed. A triangular fuzzy number can be defined
by a triplet = (a, m, b). The membership function is
with a m b. The triangular fuzzy number can be
denoted by a triplet = (a, m, b). The parameter m
gives the maximal grade of fA (X ), i.e. fA (m) = 1, it is
the most probable value of the evaluation data. a and
m are the lower and upper bounds of the available area
for the evaluation data.
For a given in the the interval [0, 1], the arithmetic operations of fuzzy numbers can be defined by
means of -cut operations according to the extension
principle(Zadeh 1965):
FTA OF CUTTERHEAD FAILURE
3.1 Fault tree model for the risk of composite EPB

shield cutterhead failure
The main purpose of the fault tree analysis is to find
out all failure modes of the system and the event with
a rather large failure probability. After weak sections
have been enhanced, occurrence probabilities of these
accidents are reduced so that the system reliability is
improved. For researching the cutterhead failure risk
of composite EPB shield, there were some supposition
in the analysis as follows:
1. It didnt take account of the disadvantageous influence that the engineering construction brought to
the surroundings building, road surface, underground pipeline etc.
2. Drag bits and scrapers would often produce normal
wear during the shield advance, so their abrasion
was not considered.
According to the above investigation accidents data,
the cutterhead failure risk of composite EPB shield in
mixed face ground was set by means of the fuzzy FTA
method, shown in Figure 1.
The systemic risk probability of cutterhead is analyzed based on the established fault tree model. Firstly,
the minimum basic event sets which cause the main
event occurrence is solved, i.e. the minimum cut set
(MCS for short) of fault tree. Each MCS corresponds
to one accident type, and there are several MCSs with
different occurrence probability of one fault tree. The
MCS with the maximum occurrence probability is the
most probably potential factor which may cause accident. Boolean algorithm is relatively simple in solving
the MCS. The fault tree in Figure 1 is obtained from
the Boolean algorithm as it shown in Eq. (6).
From the result obtained above, the top event T is the
union of 28 sets which are the MCS of the fault tree, i.e.
{X1 X7 }, {X2 X7 }, {X3 X7 }, {X3 X11 }, {X3 X16 }, {X3 X18 },
{X3 X24 }, {X3 X25 }, {X4 X11 }, {X4 X12 }, {X4 X13 },
{X4 X14 }, {X4 X18 }, {X4 X24 }, {X4 X26 }, {X22 X23 }, {X5 },
{X6 }, {X8 }, {X9 }, {X10 }, {X15 }, {X17 }, {X19 }, {X20 },
{X21 }, {X27 }, {X28 }, which correspond to 28 accident
modes.
3.2 Quantitative analysis
Quantitative analysis includes evaluation of failure
probability of the top event and important analysis of
the basic events. In practice, the occurrence probability of a top event (PT) is obtained using approximate
probability formula of independent events as shown in
Eq.(7)
Where P(Mi ) is the occurrence probability of the

ith MCS. For example, P(Mi ) is the occurrence
596
Figure 1. Fault tree of the cutterhead failure risk of composite EPB shield.
597
probability of the first MCS {X1 X7 }, which depends

on the probability multiplication of basic events X1
and X7 .
Probabilities of the basic events must be known
in advance, in order to evaluate failure probability
of the top event and important analysis of the basic
events. Expert elicitation and fuzzy set theory will
be used to get the probabilities of the basic events
in this paper. Because the experts cannot exactly
evaluate the probability of events, and sometimes
some of the events are vague, the experts tend to
apply natural linguistic expression, such as Impossible, Infrequent, Occasional, Possible and Frequent,
to describe the probability of events. According to the
Guidelines of Risk Management for Metro Tunnelling
and Underground Engineering Works (2007), it ranks
the occurrence probability of risk into 5 class, shown in
Table 1. Conventional mathematical ways cannot handle natural linguistic expression efficiently because of
its vagueness. Therefore, fuzzy set theory is used to
cope with it. According to the triangular fuzzy number
discussed above, it is assumed herein
The occurrence probabilities of random basic events

using Eq.(8) and Eq.(9), and their fuzzy probabilities
are shown in Table 2.
Figure 2 shows that the fuzzy probability of the top
event can be expressed as triangular fuzzy numbers and
the parameters are (0.9925, 0.9960, 0.9982). The corresponding occurrence probability is 0.9925 0.9982,
however, the most possible probability is equal to
0.9960 with a membership grade equal to 1.
3.3
Sensitivity analysis
The main basic events affecting the occurrence probability of the top event can be determined and some
Figure 2. Possibility distribution of cutterhead failure probability of composite EPB shield.

Table 1. The classification of risk occurring probability.
Rank
Accident description
Interval probability
One
Two
Three
Four
Five
Impossible
Infrequent
Occasional
Possible
Frequent
P < 0.01%
0.01% P < 0.1%
0.1% P < 1%
1% P < 10%
P 10%
effective measures are verified by sensitivity analysis

to reduce occurrence probability of the basic events
and the top event. According to the fuzzy number
model defined in this paper, sensitivity evaluation
index Vi is simply defined in Eq.(10) (Chen and Zhang
2002)
Where g is the occurrence probability of the top

event, xi is the average occurrence probability of the
basic event x.
If Vi Vj , it is more effective to minimize the occurrence probability of the top event by reducing the
occurrence probability of the event i rather than the
event j.
598
Table 2.
tree.
1. For lowering the risks X5 (badness geology), X8

(soft and sticky geology), X21 (alternated with soft
and rigidity rock terrane) influences on the cutterhead failure, it should strengthen to run the
geology forecast, accurately certain the position of
the badness geology and its distribute, and adopt
corresponding measures in advance.
2. For lowering the risk X6 (irrationality type of cutterhead) influence on the cutterhead failure, it should
accord to the geology and hydrology condition,
structural design, construction advance request etc.
factor, choose much adaptability cutterhead, and
make an adequacy adjustment of shield installation
under concrete conditions during construction.
3. For lowering the risks X4 (exceeding allowed abrasion), X7 (misgovern construction) influences on
the cutterhead failure, it should choose reasonable advance model and parameters, continuously
accumulate the experience, and reduce man-made
breakage; installing the cutter wear monitor system, it can accurately obtain the information of the
cutter wear, then adopt corresponding measures;
enhancing cutters replacing rate and quality.
Fuzzy probability of the basic event in the fault

Fuzzy probability value
Basic
event
Sensitivity
index
X1
X2
X3
X4
X5
X6
X7
X8
X9
X10
X11
X12
X13
X14
X15
X16
X17
X18
X19
X20
X21
X22
X23
X24
X25
X26
X27
X28
0.000095
0.000475
0.095
0.8075
0.0855
0.0475
0.0095
0.76
0.000665
0.000475
0.0057
0.00076
0.095
0.0038
0.076
0.057
0.00855
0.00076
0.000665
0.000855
0.76
0.0855
0.00665
0.0057
0.0038
0.00057
0.00038
0.057
0.01%
0.05%
10%
85%
9%
5%
1%
80%
0.07%
0.05%
0.6%
0.08%
10%
0.4%
8%
6%
0.9%
0.08%
0.07%
0.09%
80%
9%
0.7%
0.6%
0.4%
0.06%
0.04%
6%
0.000105
0.000525
0.105
0.8925
0.0945
0.0525
0.0105
0.84
0.000735
0.000525
0.0063
0.00084
0.105
0.0042
0.084
0.063
0.00945
0.00084
0.000735
0.000945
0.84
0.0945
0.00735
0.0063
0.0042
0.00063
0.00042
0.063
1.0043e-6
5.0216e-6
0.0083
0.9532
0.0803
0.0502
0.0010
0.8035
7.0302e-4
5.0216e-4
0.0057
6.8294e-4
0.0854
0.0034
0.0803
0.006
0.009
8.0345e-5
7.0302e-4
9.0389e-4
0.8035
6.3272e-4
6.3272e-4
0.0051
4.0173e-4
5.1e-4
4.0173e-4
0.0603
From risk analysis of composite EPB shield cutterhead

failure in the mixed face ground based on fuzzy FTA,
the conclusions are as follows:
Sensitivity indices of all basic events are obtained

by sensitivity analysis of the fault tree of the cutterhead
failure risk of composite EPB shield as shown in Fig.1.
The basic event in which the sensitivity index
is greater than 5% is chosen and arranged as follows: V4 = 95.32%, V8 = V21 = 80.35%, V13 = 8.54%,
V5 = 8.03%, V28 = 6.03%, V24 = 5.1%, V13 = 5.02%.
4
CONCLUSIONS
MAINLY INFLUENCE FACTOR AND

IMPROVENMENT MEASURE
According to the order result, single minimum cut set

X5 , X6 , X5 , X8 , X21 and so on easily cause the failure of cutterhead; Secondly, the basic events X4 , X7 ,
X24 which appear more times also easily cause the top
event failure. They are the weakness parts of the system and the main risk factors arousing the failure of the
shield cutterhead in the mixed face ground. Therefore,
during the tunnel construction, it aims at surveying
and managing the basic events which greatly influents
the top event occurrence to lower the risk accidents of
cutterhead failure.
1. Having a directly view and simple character, the

FTA is a valid method for analyzing the failure risk
of composite EPB shield cutterhead in mixed face
ground.
2. It totally considered 28 basic events for the failure
risk fault tree of composite EPB shield cutterhead. Through the fuzzy fault tree calculation, it
can definite the weakness parts of cutterhead failure, confirm the key factors of risk occurrence,
and make an order for the importance of various
influence factors.
3. Through fuzzy FTA of the cutterhead failure, it
made sure the mostly reason and mechanism of
leading to the failure risk of composite EPB shield
cutterhead, brought forward the improved measures
and suggestion. So it gave some useful conference
for preventing or reducing the failure risk of composite EPB shield cutterhead during construction.
REFERENCES
Cao, H.L., Lv, C.T. & Li, J.X. 2004. Failure analysis for cutters
and its prevention measures. Tunnel Construction 24(6):
910.
Evans, J.R. 2001. Itroduction to Simulation and RiskAnalysis.
2/e, Prentice Hall Business Publishing.
599
Furuta, H. & Shiraishi, N. 1984. Fuzzy importance in Fault

tree analysis. J.Fuzzy sets and systems 12: 205213.
Guo, L. 2006. Fault elimination of mixed face shield during construction. Modern Tunneling Technology (supplement): 436439.
Meng, X.J. 2004. Causation analysis and resolve measures
for the ordinary fault (damage) of the mixed EPB shield
cutterhead. Tunnel Construction 24(2): 6166, 73.
Singer, D. 1999. Fuzzy set approach to fault tree and
reliability analysis. J.Fuzzy sets and systems 34: 145155.
Tanaka, H., Fan, L.T. & Taguchi, T. 1983. Fault tree analysis by fuzzy probability. J. IEEE Trans on Reliability 32:
453457.
Tongji University 2007. Guidelines of Risk Management for
Metro. Tunnelling and Underground Engineering Works.
2007.
Zadeh, L.A. 1965. Fuzzy sets Information and Control 8:

338353.
Zhang, G.D. & Lu, Y.X. 1990. Analysis and design
of reliability and maintenance of system. Beijing:
Beijing Aeronautics and Astronautics University Press:
120125.
Zhao, J., Gong, Q.M. & Eisensten, Z. 2007. Tunnelling
through a frequently changing and mixed ground: A case
history in Singapore. Tunnelling and Underground Space
Technology 22: 388400.
Zhu, W.B., Ju, S.J. etc. 2006. Shield tunnelling technology in
mixed face ground conditions. Beijing: China Science and
Technology Press.
600
Risk assessment on environmental impact in Xizang Road Tunnel

C.P. Yao, H.W. Huang & Q.F. Hu
ABSTRACT: Nowadays, with the increase in the demands for environmental protection, the assessment of
construction on the environmental impact has become a new topic. Based on the Risk Management Software
(TRM 1.0) which is developed by Tongji University, the authors investigate the surrounding environment condition and construction process, calculate the probability of hazard risks and its loss, and provide the ranks of
hazard risks and the corresponding treatment on risks. It may give some advice and suggestions on underground
construction.
INTRODUCTION
risk reduction measures. It may give some advice and

suggestions on underground construction.
Because of its unique unseen character, isolation from

nature and complexity, the construction of underground works is more difficult than the process of
general buildings. In the age of high demands of
environmental protection, the environmental impact
caused by underground works construction is attracting increasing attention. The paper is for the risk
assessment on environmental impact caused by the
construction of Shanghai Xizang South Road Tunnel from the point view of risk. The authors make
quantitative analysis on the impact of construction
on surrounding buildings, the impact on surrounding
roads and the impact on surrounding pipelines. At last,
the ranks of hazard risks are provides as well as the
PROJECT PROFILE
Xizang South Road Tunnel is a planning river-crossing

tunnel in Shanghai Expo. The main line of the tunnel is
from the intersection of Puxi Xizang South Road and
Zhongshan South Road to the intersection of Pudong
Binzhou Road and Yunlian Road, including the intersection of Binzhou Road and Pudong South Road. The
total project is 2673 meters long. Figure 1 shows the
road tunnel plan.
The tunnel is advanced by Slurry Balance Shield
machine of 11.36-meter diameter. Project consists
Figure 1. Xizang south road tunnel plan.
601
the construction process, and then classify, collating those parameters affecting greatly.
2 Risk analysis: Calculate the probability and consequence of risk factors.
3 Risk evaluation: According to some certain criteria,
evaluate the risk factors.
Risk assessment on
environmental impact
Impact on
surrounding
buildings
Impact on
surrounding
roads
Impact on
surrounding
pipelines
Figure 2. Risk identification map of environmental impact.
of the approach of Puxi, Puxi rectangular tunnel

(including wells), the circular tunnel, Pudong rectangular tunnel (including wells) and Pudong approach.
Along the setting location, there are great population,
many buildings, complex pipelines and complex traffic systems. Also, the tunnel will up-cross Line 7 and
down-cross Line 8, both of which are components of
Shanghai Railway Transit System. Moreover, the tunnel will intersect with magnetic train line which will
be constructed soon. In the later part, the authors will
assess the risk of environmental impact caused by construction, including impact on surrounding buildings,
on surrounding roads and on surrounding pipelines. It
is shown in Figure 2.
3
RISK ASSESSMENT ON ENVIRONMENTAL

IMPACT
3.1
Where, risk evaluation is to use Experts Investigation Method (EIM) and Confidence Index Method
(CIM), to analyze the probability and consequence
of risk factors, and to gain the ranks of risks. EIM
is a kind of gathering information method. We send
investigation forms concerning with the risk factors
in engineering to experts, professors, senior consulters and so on. The CIM is just an index to show their
confidence when doing the judgment. The data can
be put into the database of TRM 1.0, which is developed by Tongji University (Huang, et al 2006b). The
detailed data-input procedure can refer to the reference Huang, et al 2006b. Table 1 shows part of the risk
investigation form.
In this paper, the applied risk evaluation criteria is Guidelines for Tunnelling Risk Management
enacted by the International Tunnel Association in
2002 and Guidelines of Risk Management for Metro
Tunnelling and Underground Engineering Works
enacted by Tongji University (International Tunnel
Association, 2002 & Huang, et al 2006a). They are
shown in Tables 2, 3, 4 and 5.
3.2 Risk assessment on environmental impact
on surrounding buildings
Risk assessment process
Risk assessment is usually divided into three steps

(Guo, 1986 & Chen, 2004):
1 Risk identification: Analyze all potential risk factors which may influence the environment during
3.2.1 Risk identification

The possible risk hazards of surrounding buildings
caused by construction can be illustrated in Figure 3
(Yao, et al 2006).
Table 1. Risk investigation form (part of it).

Environmental assessment
Occurrence probability
Loss
Risk hazard
Probability rank
Loss rank
Table 2.
Risk-inducing factor
Confidence index
Confidence index
Risk occurrence probability ranks.
Ranks
Occurrence
probability
Impossible
P < 0.01%
Seldom
0.01% P < 0.1%
Occasional
0.1% P < 1%
Possible
1% P < 10%
Frequent
P 10%
Table 3.
Loss ranks of risk hazards .
Ranks
Details
Ignored
Considered
Serious
Very serious
Disastrous
Here, the criterion on rank of loss differs from country to country. In China, the loss is disastrous when 5 lives lost.
602
3.2.2 Risk analysis

The overall sinking of buildings is caused by the
even settlement of ground. Uneven settlement can
induce the tilting and cracking of buildings. During the
construction process, many factors can cause ground
settlement (Yao & Huang, 2007). For example, the
seepage of shield and pit, pit collapse, the improper
earth pressure of shield and support damage. Such
factors may also be the cause of building damage. If
this happens, it will cause great social influence and
economic loss.
Table 4.
The passing buildings of the tunnel can be shown

in Table 6, Appendix 1 & 2, respectively.
We may see, the buildings are very dense, therefore
the risk is high. It is very easy to tilt or crack for the
buildings during the construction.
3.2.3 Risk evaluation
By sending out investigation forms, and using theTRM
1.0, the authors get the risk ranks of these risk hazards. The results are shown in Table 7, from which we
may see buildings crack is more risky and is highly
probable to happen.
Risk evaluation matrix.

Loss
Risk
Occurrence
probability
A: P < 0.01%
B: 0.01% P < 0.1%
C: 0.1% P < 1%
D: 1% P < 10%
E: P 10%
1. Ignored
2. Considered
3. Serious
4. Very serious
5. Disastrous
1A
1B
1C
1D
1E
2A
2B
2C
2D
2E
3A
3B
3C
3D
3E
4A
4B
4C
4D
4E
5A
5B
5C
5D
5E
Table 5.
Risk acceptance criteria.
Ranks
Risk
Acceptance criteria
Measures
I
II
III
IV
V
1A,2A,1B,1C
3A,2B,3B,2C,1D,1E
4A,5A,4B,3C,2D,2E
5B,4C,5C,3D,4D,3E
5D,4E,5E
Ignored
Allowable
Accepted
Unaccepted
Cannot be accepted
No need for management and supervision

Call for attention, need regular management and supervision
Call for great attention, need prevention, and monitoring measures
Need the decision of policy-maker, need controlling measures
Immediately stop, need alternative plans
Table 6.
Surrounding buildings in Puxi and Pudong District.
District
Puxi
district
Buildings
Manufacturing
bureau
Xizang south road
South station road
Jiangnan shipyard
Pudong
district
Shanghai hangbiao
factory
Nanchuan factory
Shangnan road
Pudong south
road-Yaohua road
Binzhou road
6 buildings with 5 6 floors, 2 buildings with 4 floors,

2 buildings with 1 floor, 2 buildings with 3 floors, 1 building with 10 floors, 1 building
with 7 floors
Many buildings with 3 7 floors, 2 buildings with 15 floors, 1 building with 12 floors,
1 building with 10 floors, several buildings with 1 floor, 1 building with 19 floors,
1 building with 28 floors (reserved for Shanghai Expo)
1 building with 16 floors, 1 building with 5 floors, 2 buildings with 6 floors, 1 building
with 7 floors
Three-floor defense project, 1 building with 6 floors, distribution substation with
5 floors, 3 docks
Floating pier, floodwall, 3 steel pipe piles, several buildings with 3 6 floors
Plant with 1 3 floors, 1 building with 4 floors, 2 buildings with 6 floors
Several buildings with 1 3 floors, 1 building with 15 floors, 1 building with 13 floors,
a library with 4 floor
Several buildings with high floors, buildings with 6 floors
10 buildings with 4 6 floors
603
Table 8.
Risk assessment of impact on surrounding roads.
Impact on surrounding buildings

Risk hazard
Probability
Loss
Rank
Road
damage
II
C
C
2
3
II
III
III
Building
crack
Building tilt
Overall
sinking of
buildings
No.
Surface
subsidence
Road uplift
Road
fracture
Traffic jam
Figure 3. Risk identification of impact on surrounding

buildings.
Table 7. Risk assessment of impact on surrounding
buildings.
No.
Risk hazard
Probability
Loss
Ranks
Overall sinking
of buildings
Building tilt
Building crack
II
B
C
3
3
II
III
2
3
Pipeline damage
Gas
pipeline
damage
Water
pipeline
damage
Sewage
pipeline
damage
Electricity
pipeline
damage
Telecom
pipeline
damage

pipelines.
Table 9. Risk assessment of impact on surrounding
pipelines.
Road risk
Road damage
Surface
subsidence
Road uplift
Traffic jam
Road
fracture

roads.
3.3
Risk assessment on environmental impact on

surrounding roads

The possible risk hazards of surrounding roads caused
by construction can be illustrated in Figure 4.
3.3.2 Risk analysis
The traffic situation along the works is extremely complex. The tunnel is the link between Puxi and Pudong
district, along with Nanpu Bridge, Lupu Bridge and
Da-pu road tunnel. Once the road is damaged, causing
traffic jam, the consequences will be serious.
There are many roads along, but because of the limit
of the pages, the figure of roads is omitted.The analysis
is just the same as above.
1.0, the authors get the risk ranks of these risk hazards.
No. Risk hazard
Probability Loss Rank
1
2
3
4
5
B
B
C
B
C
Gas pipeline damage

Water pipeline damage
Sewage pipeline damage
Telecom pipeline damage
Electricity pipeline
damage
3
4
2
3
3
II
III
II
II
III
The results are shown in Table 8, from which we may

see traffic jam and road fracture are more risky and
are highly probable to happen.
3.4 Risk assessment on environmental impact on
surrounding pipelines
The possible risk hazards of surrounding pipes caused
by construction can be illustrated in Figure 5.
3.4.2 Risk analysis
There are many pipelines along, but because of the
limit of the pages, the figure of pipelines is omitted.
The analysis is just the same as above.
1.0, the authors get the risk ranks of these risk hazards.
The results are shown in Table 9, from which we may
see water pipeline damage and electricity line damage
are more risky.
604
Table 10.
3 Remove some load pushed on the pipelines.

4 Take informational construction. Monitor the
pipelines frequently.
Risk assessment of environmental impact.
No.
Risk hazard
Rank
1
2
3
Impact on surrounding buildings

Impact on surrounding road
Impact on surrounding pipelines
III
III
III
CONCLUSION
From the study and research on this project, we may

get some useful conclusions:
3.5
Summary
1 The whole risk rank for Xizang South Road Tunnels environmental impact is III. It should be paid
more attention. When the tunnel is constructed,
the workers may strictly control ground settlement.
Information construction is recommended.
2 There are plenty of methods to evaluate risk. It
needs to discuss deeply which method is adopted.
In this paper, Experts Investigation Method (EIM)
and Confidence Index Method (CIM) are applied,
and they are applicable.
3 The risk rank of environmental impact is obtained,
which may give advice on project decision, bidding
and insurance.
4 Risk management is a dynamic progress. With the
development of the project, the risk rank may be
changed. The authors should follow the project, and
realize the dynamic risk management.
The whole risk rank for environmental impact is III.

The result is shown in Table 10. It should be paid more
attention.
Road tunnel construction on surrounding buildings
impact was mainly due to surface subsidence during the process, causing buildings sinking, tilting or
cracking. The main treatments are:
1 Grout timely, and make sure the shield is well
sealed.
2 Control the earth pressure during the digging to
prevent the face instability, a surface subsidence
and uplift.
3 Protect the important buildings. If it is necessary,
consolidate the foundation.
4 All kinds of anti-leakage measures are needed.
Pay attention to foundation strengthening process,
otherwise it will most prone to landslides.
5 Focus on the steps of construction during the
foundation engineering. Control every step of the
excavation depth and slope, including good support
and the timely installation;
The measures about road and traffic involve with
following instructions.
1 Strictly control the quality of temporary road construction, to ensure road safety.
2 Control the load of vehicles travelling on the road
to prevent overweight and crushing road.
3 Detailedly understand the traffic flow near the
project before starting the construction. Disperse
traffic flow to avoid jam.
There are many large diameter pipelines near the
project, and once the pipelines are destroyed, it will
not only have huge economic losses, but also resulted
in a very bad social influence. Therefore we should
take protective measures against dangerous pipelines.
1 Since there may be some errors occurring when
planting the pipelines, which may induce incorrect position of pipelines. So before construction,
positions of important pipelines must be verified.
2 Protect important pipelines by consolidating earth
or isolation method.
REFERENCES
Chen, L. 2004. Risk Analysis and Assessment During Construction of Soft Soil Shield Tunnel in Urban Area. PhD
Thesis of Tongji University.
Guo, Z. W. 1986. Risk analysis and decision. Beijing:
Machinery Industry Press.
Huang, H. W. et al 2006a. Guidelines of Risk Management for
Metro Tunnelling and Underground Engineering Works.
Tongji University.
Huang, H. W. et al. 2006b. Risk Management Software
(TRM1.0) Based on Risk Databas for Shield Tunnelling.
Chinese Journal of Underground Space and Engineering.
Vol. 2: 3641.
International Tunnel Association. 2002. Working Group
No. 2. Guidelines for tunnelling risk management.
Balkema.
Yao, C. P., Huang, H. W. & Hu, Q. F. 2006. The Study on
Choosing Construction Methods of Underground Works
Based on The Risk Analysis. Modern Tunnelling Technology. Vol. 43: 581585.
Yao, C. P. & Huang, H. W. 2007. Risk analysis on the
construction of connection tunnels in Shanghai Yangtze
River Tunnel. Underground Construction and Risk
Prevention Proceedings of the 3rd International Symposium on Tunnelling-Shanghai 2007: 99104. Tongji
University Press.
605
Appendix 1. Surrounding buildings in Puxi district.
Appendix 2. Surrounding buildings in Pudong district.
606
Risk analysis and fuzzy comprehensive assessment on construction

of shield tunnel in Shanghai metro line
H.B. Zhou, H. Yao & W.J. Gao
Shanghai Jianke Project Management Co., Ltd., Shanghai, P.R. China
ABSTRACT: Based on risk management of metro line in Shanghai the risk analysis and fuzzy comprehensive
assessment model was introduced to the very long metro line project risk assessment of shield tunnel construction.
Firstly, Work Breakdown Structure method was brought forward to break down the shield tunnel construction
works according to the features of the engineering geology and surrounding building and underground public
pipes, the fault tree method was used to identify the shield tunnel constructions risk events and risk factors and
the risk list can be established. Secondly, the fuzzy comprehensive assessment model was presented to assess the
construction risk in view of the complexity of underground space and difficulty of quantifying the evaluation
parameters, and the sub-projects risk level and the project overall risk level can be calculated . Finally, a case
about very long metro line project in Shanghai was studied, the assessment results is consistent with reality.
INTRODUCTION
Eleven metro lines projects with the total length of

more than 400 kilometers will be constructed in
Shanghai before 2010, the construction of metro systems in Shanghai is in Large-scale construction now.
Metro systems can effectively decrease traffic congestion, at the same time, due to the construction of
metro mainly taken underground with many uncertainties as well as limited time, many serious accidents
has happened such as the construction accident happened in Shanghai Rail Transit Line 4 in July 2003
and the collapse in Guangzhou Metro Line in 2004.
These accidents not only brought enormous loss to the
country, but also exerted a negative influence on the
public. So, the risk assessment on the shield construction has vital significance to insure the project proceed
smoothly (Mao 2004, Su 2004).
In this paper, a new risk analysis and assessment
method during very long shield tunnel construction
was put forward. Firstly the risk analysis method was
introduced, which analyzed metro construction risk
combining the WBS method with fault tree method
because that the very long metro constructions risk
was very complex, in the risk analysis process, the very
long metro line project was classified according to the
features of the engineering geology and surrounding
building. Then the Work Breakdown Structure (WBS)
method was brought forward to break down the shield
tunnel construction works. The fault tree method
was used to identify the shield tunnel constructions
risk events and risk factors, so the risk list can be

established. Secondly, a new fuzzy comprehensive
assessment model was presented to assess the construction risk in view of the complexity of underground
space and difficulty of quantifying the evaluation
parameters, which was proposed based on the fuzzy
mathematical theory and risk matrix method proposed
by International Tunneling Association in 2004. At last
a case about one very long metro line construction
risk assessment was introduced to demonstrate the risk
analysis and risk assessment method. The assessment
results were consistent with reality. All the risk analysis and assessment ideas and method given about very
long metro line project are of great practical value for
similar mega engineering construction risk assessment
in the future.
2
RISK ANALYSIS METHOD RESEARCH
Now, there were many risk analysis methods such as

WBS-RBS method, fault tree method and so on. If the
WBS-RBS method was used to analyze the risk factors of construction of very long metro line project, one
very huge matrix would be established to identify the
risk and the method could not distinguish relationship
between the risk factors. The fault tree method also
could not solve the problems. So a new risk analysis
method that combined the WBS method with fault tree
method was proposed. Firstly, the very long metro line
project was classified according to the features of the
607
engineering geology and surrounding building. Then

the Work Breakdown Structure method was brought
forward to break down the shield tunnel construction
works. The fault tree method was used to identify
the shield tunnel constructions risk events and risk
factors, so the risk list can be established.
Table 1.
2.1 Work breakdown structure method

In order to analyze the risk, The work breakdown structure (WBS) method was brought forward to break
down the metro construction works and to analysis
the risk events and risk factors.
2.1.1 The first grade structure breakdown
The metro line was divided into several sections to
analyze risk according to the features of the engineering geology, surrounding building, underground
public pipes and traffic conditions.
2.2
Degree
Estimate
value
Slight
Medium
1
2
Serious
Significant
Disastrous
Table 2.
2.1.2 The second grade structure breakdown

In order to analyze the risk of every section comprehensively during the construction, the activities of
each section are divided further into portal-out, shield
driving, segment works, grouting construction, shield
machine, portal in and contact channel.
Estimation method of the risk effect consequence.
Explanation
The loss is not obvious
The loss is few (minimum in
100,000 RMB)
The loss is compensable
(minimum in 1,000,000 RMB)
The loss is great but compensable
(minimum in 10,000,000 RMB)
The loss is too enormous to
compensate (above 10,000,000
RMB)
Estimation of the risk probability.
Probability
Estimate value
Frequency
Rare
occasional
Possible
anticipated
frequent
1
2
3
4
5
<0.0003
0.00030.003
0.0030.03
0.030.3
>0.3
Risk identification
The fault-tree method was used to identify the risk

events and risk factors based on breaking down the
works (Loosemore, Raftery, Reilly etal 2006), then the
risk event list can be established.
the indicators importance (Fang & Wang 1997, Shao

2004).
3.2 Calculating the risk events degree of

membership
FUZZY ASSESSMENT METHOD

RESEARCH
There have been limited attempts to exploit fuzzy logic

within the construction risk management domain. Previous approaches to the use of fuzzy logic within
construction risk management have proved to be either
too simplistic for use in the real world, or have been
very specific in their approach, targeting a particular
area of construction on which to act or concentrating
on specific types of risks.
In this paper, a new fuzzy comprehensive evaluation
model used to the risk assessment was proposed, which
attempted to solve previous problem. The fuzzy comprehensive evaluation model was established based on
the fuzzy mathematical theory and risk matrix method
proposed by International Tunneling Association in
2004. It can calculate the evaluating indicators degree
of membership and the weights, and the model can
make sure the computed result is more objective.
3.1
Calculating the indicators weights
The indicators weights were determined by the Analytic Hierarchy Process (AHP) method by comparing
Firstly, the expert scoring method was used to estimate

each risk events values of effect consequence C and
possibility P, then the product of P multiply C was led
to the risk ranks membership function and the risk
level degree of membership to the every risk event are
determined.
3.2.1 Determination of the risk event effect
consequence C
The effect consequence C of risk event is determined
according to the shield tunneling constructions actual
situation (Mao 2004). The estimate method is shown
on table 1.
3.2.2 Determination of the risk events possibility
The risk events occurrence probability is ranked into
five levels; estimate method is shown on table 2.
3.2.3 Calculating the degree of membership
3.2.3.1 Determination of membership function
In 2004, the International Tunnel Association put forward the risk matrix theory used to assessment risk,
in this theory; the risk was classified four rank. In
608
degree of membership
1.2
level 1
level 3
level 2
level 4
1
0.8
0.6
0.4
0.2
0
10 12 14 16 18 20 22 24 26 28 30
PC
Figure 1. Membership function for risk.
Figure 2. The chart of the metro route map.
Table 3. The membership function.

Risk level
value C into the risk rank membership function, then

the corresponding membership degree of the risk
events are determined.
Membership function
level 1
CASE STUDY OF RISK ANALYSIS AND

ASSESSMENT
One length of the metro line is about 34.243 kilometers

in all in Shanghai area, which passes through five district of Shanghai including Baoshan district, Putuo
district, Jinan district, Xuhui district and Pudong new
district from north-west to south-east. The metro lines
route map is shown on figure 2.
level 2
4.1 Geological conditions
level 3
4.1.1 Engineering geological conditions

The on-site soils within 65 m depth are the Quaternary
sediments, which are mainly composed of saturated
soft clay, silty clay and sand. The main soil profiles
of the construction of shield tunnel are saturated soft
clay and silty clay.
level 4
this paper, the risk rank also was classified into four
levels. The membership function was the L-R fuzzy
function which used in project very widely, as shown
on Equation 1, to be calculated easily, and according to the feature of shield tunnel construction risk
assessment, the function was designed to trapezoidal
function, as shown on figure 1 and table 3. (Carr &
Tah 2001, Wang & Li 1996,Yao & zhou 2007).
4.1.2 The hydrogeological conditions

There are three kinds of groundwater in the soils,
including the upper phreatic water mean embedded
depth is 0.5 0.7 m under the ground surface, the middle feeble confined water mean embedded depth is
36 m under the ground surface in the 2-2 sub-layer of
fifth layer soils and lower confined water mean embedded depth is 414 m in the seventh layer soil. The water
elevation of groundwater was changed with the change
of season, climate and tide.
4.2 Risk analysis
3.2.3.2
Determination of membership degree value

of risk events
Take the product of occurrence probability estimated
value P and multiply consequence effect estimated
The metro line was divided into 12 sections to analyze risk according to the features of the engineering
geology, surrounding building, underground public
pipes and traffic condition, as shown on table 4
then the shield tunneling constructions work structure was broke down, so the risk list was established
609
Table 4. The list of divided evaluating sections of metro line.

Category
surrounding environments
Mainly traversing soils of shiled
The 1st section
sub-center of the city and the complex of surrounding

environments is common
sub-center of the city and surrounding environment is
relatively complex
center of the city and traverse under the North Zhongshan road
center of the city and surrounding environment is complex
center of the city and traverses through Suzhou River
center of the city and traverses metro No. 1 line and metro
No. 2 line closely
center of the city and there are piles foundations intruding in
the metro tunnel
cross the Huangpu river and construction environment is very
complex
Traverse under or beside piles foundation and surrounding
environment is complex
the complex of surrounding environment is common
traverses the river below and the complex of
surrounding environments is common
traverses under metro No. 2 line
soft clay and silty clay
the 2nd section

the 3rd section
the 4th section
the 5th Section
the 6th Section
the 7th Section
the 8th Section
the 9th Section
The 10th section
The 11th section
The 12th section

gray soft clay, clay and sandy silt
gray soft clay and clay
Shield tunneling (the 8th section)
Too much distortion

of the flood prevention
wall
A departure from the

designed axes of
tunneling
Construction risk
through the middle
cession
Sinking
Sealed equipment
leaking
failure of
working face
grouting is not
timely
pushing force is
unbalance
change of
geotechnical
condition
driving axis not

timely corrective
over driving or
less driving
excessive
measurement error
Has not been able to

sink promptly puts the
arriving
position deviation
oversized
shield standstill
soft ground
failure of
workingface
sinkhole
seal oil is
unqualified
seal oils
quantity is short
sealed equipment
loss flexibility
mechanical failure
construction
parameters of
improper
failure of
working face
Soils pouring out,

quicksand
Collapse
Survey unclear
change of
geotechnical
condition
Meeting obstructions
Figure 3. Shield tunneling constructions risk events and risk factors.
by using fault tree method to analyze shield tunneling construction risk. As an example, the result of
risk identification about shield driving construction
of the 8th kind of sector project (under Huangpu river
section) was shown on figure 3.
4.2.2 Membership function determining

Taking the product of possibility P of the risk event and
consequences C into the membership function, each
risk events degree of membership was determined.
The judging matrix R was also formatted and the value
was as follows:
4.2.1 Weights calculation

The AHP method was used to determine the weight of
each risk event. The estimation matrix was formatted
based on the mutual comparison of important degree
among the risk events.
Operating on the estimation matrix, the vector
corresponding valued W of shield tunneling was as
follows:
W = [0.0460.2070.2700.0260.0460.0780.2070.12]
4.2.3 Fuzzy comprehensive assessments

Taking weight vector W and judging matrix R
into Fuzzy Comprehensive Assessment, the whole
610
Table 5. The point of each evaluated index.
Shield
tunneling
construction
risk events
meeting obstructions
collapse
soils pouring out, quicksand
sealed equipment leaking
sinking of the shield
a departure from the designed
axes of tunneling
construction risk through the
middle caisson
too much distortion of the flood
prevention wall
2
1
3
3
3
4
3
5
3
2
3
2
Table 7.
line.
Risk level calculated of each section of the metro
category
risk level
the 1st section

the 2nd section
the 3rd section
the 4th section
the 5th section
the 6th section
the 7th section
the 8th section
the 9th section
the 10th section
the 11th section
the 12th section
total risk level
Level 2
Level 3
Level 4
Level 4
Level 3
Level 3
Level 4
Level 4
Level 3
Level 2
Level 2
Level 4
Level 3
Table 6. the shield tunneling construction risk events risk

level of the 8th section.
construction
work of the risk
8th section
level
shield
tunneling
level 3
risk events
risk
level
meeting obstructions
level 2
collapse
level 3
soils pouring out,
level 3
quicksand
sealed equipment leaking
level 2
sinking
level 3
construction risk through
level 3
the middle cession
a departure from the
level 4
designed axes of tunneling
too much distortion of the
level 3
flood prevention wall
leaking. In the meanwhile, risks of other construction

activity during shield tunnel construction in 8th section
were calculated; the result of risk assessment about the
8th section is as shown on table 6. In the end, risk level
of the 8th construction section could be determined by
the calculating upwards.
The risk level of other 11 construction section could
be determined by the same method and the result is
shown on table 7. From the table 7, it can be concluded the whole risk level of this ultra-length metro
line project is 3. So certain measures must be taken to
reduce the risk based on the risk acceptance principle.
CONCLUSIONS
In accordance with the study on the risk analysis

and assessment method of shield tunneling construction of very long metro line project and case study,
conclusions were given as follows:
evaluation vector B about the tunneling construction

was determined. The value was as follows:
Based on the maximal membership degree principle, the tunneling construction risk rank about shield
tunneling of the 8th section is level 3, in which the
highest risk is the construction through the middle
caisson and the lowest risk is construction of meeting underground obstructions and sealed equipment
611
1 A new risk analysis method that combined the WBS

method with fault tree method was proposed, in
which classified the very long metro line project
according to the features of the engineering geology and surrounding building, then broke down
the shield tunnel construction works, and used
fault tree method to identify the shield tunnel constructions risk events and risk factors. The method
not only can identify the risk events and risk factors more comprehensively, but also distinguish
the relationship between each risk events or risk
factors.
2 The fuzzy comprehensive risk assessment model
could be used to solve the problem of non-accuracy,
quantify the subjective concept, convenient mathematical processing and reduce the difference
caused by subjective judgment.
3 The fuzzy comprehensive risk assessment model

proposed in this paper could be operated upward
from the bottom layers of risk structure up to
obtain the overall evaluation results of the whole
shield tunnel construction project. The result of risk
assessment is very consistent with the practical construction conditions. Therefore, the method above
mentioned is very practical and credible.
REFERENCES
Carr, V. & Tah, J.H.M. 2001. A fuzzy approach to construction project risk assessment and analysis: construction
project risk management system.Advances in Engineering
Software (32): 847857.
Fang, S.C. & Wang, D.W. 1997. Fuzzy mathematics and fuzzy
optimization. Beijing : Science Publishing Company.
Loosemore, M., Raftery, J., Reilly, C. & Higgon, D. 2006. Risk

Management in Projects. London :Tayor and Francis.
Mao, R. 2004. Risk management on underground engineering. International Proseminar Memoir about Engineering
Managemengt(11): 9498.
Shao, R.Q. 2004. A multi-level fuzzy synthetic evaluation on investment programs in shipping. Transportation
Engineering 144(93): 497502.
Su,Y. 2004. Seismic risk management and insurance analysis
of underground structures in soft soil. Doctor Thesis of
TongJi University: 85105.
Wang, Z.P. & Li, H.X. 1996. Fuzzy system theory and fuzzy
computer. Beijing : Science Publishing Company.
Yao, H. & Zhou, H.B. 2007. Fuzzy synthetic evaluation on
the constructing risk of EPBS during tunnelling in soft soil
area. Rock and Soil Mechanics. Vol. 28(8):17531756.
612
Theme 5: Physical and numerical modelling
Tunnel behaviour under seismic loads: Analysis by means of

uncoupled and coupled approaches
D. Boldini
Department of Chemical, Mining and Environmental Engineering, University of Bologna, Bologna, Italy
A. Amorosi
Department of Civil and Environmental Engineering, Technical University of Bari, Bari, Italy
ABSTRACT: In this paper different approaches to investigate the behaviour of tunnel under seismic loads
are presented. They include one-dimensional (1D) numerical analyses performed modelling the soil as a single
phase non-linear visco-elastic medium, the results of which are then used to evaluate the input data for selected
analytical solutions proposed in the literature (uncoupled approach), and 2D fully coupled Finite Element (FEM)
simulations adopting a visco-elastic effective stress model for the soil (coupled approach).
INTRODUCTION
The dynamic response of tunnels to seismic actions

can be assessed by means of uncoupled or coupled
approaches, depending on whether the evaluation of
the seismic wave propagation and of the corresponding
actions on the structure is undertaken in two separated
steps or in one single analysis, respectively.
In this paper the transverse seismic behaviour of
an idealised shallow tunnel in soft clay is analysed
by means of uncoupled and coupled approaches. They
include one-dimensional (1D) numerical analyses performed modelling the soil as a single phase non-linear
visco-elastic medium and 2D fully coupled Finite
Element (FEM) simulations adopting a visco-elastic
effective stress model for the soil.
In the uncoupled approach 1D visco-elastic analyses, performed using the equivalent linear scheme
implemented in the code EERA (Bardet et al. 2000),
are aimed at establishing the role of stiffness and
damping non-linearity on the free-field site response.
The results of the analyses at the tunnel depth are
then used to evaluate the input data for selected analytical solutions proposed in the literature to predict
the transverse response of the structure both for fullslip and no-slip conditions (e.g., Wang 1993). This
approach, widely used in the engineering practice,
is based on the assumptions that the free-field site
response is representative of the problem under study
and that the seismic event at the tunnel depth can be
satisfactorily represented by means of the maximum
shear strain induced in the soil and the corresponding
mobilised stiffness.
To overcome some of the limitations of the approach

described above a fully coupled Finite Element analysis can be adopted, simulating in the time domain the
soil-structure dynamic interaction during the seismic
event. This latter is in this case realistically described
by an accelerogram and the soil by an appropriate
effective stress formulation.
The constitutive assumption for the soil is a key element of this class of analyses. A linear visco-elastic
model is often adopted at the scope, for its apparent simplicity and limited number of parameters. In
fact, a number of commercial codes nowadays allow
the user to perform coupled dynamic analyses based
on linear elasticity and viscous damping, this latter
accounted for by the Rayleigh formulation. In this
case a limitation of the coupled approach is in the non
obvious selection of adequate elastic and viscous soil
parameters, which sensibly influence the results of the
analysis.
In this work a strategy to calibrate the parameters for
the visco-elastic model adopted in the FEM analyses is
proposed based on the free field soil response results
obtained in the context of the uncoupled approach.
A critical comparison between coupled and uncoupled approaches is outlined in the paper with reference
to a shallow tunnel excavated by a TBM. A realistic geotechnical characterisation is assumed for the
idealised soft clay deposit under study.
2
CASE STUDY
In the present study the acceleration time history

recorded at Kalamata (Greece) during the 13.XI.1986
615
earthquake is considered.The original seismic signal is

characterised by a duration of 29.74 s and a maximum
acceleration of 0.24 g.
The input signal adopted in this work was scaled
at 0.35 g and was filtered to prevent frequency levels
higher than 7 Hz. A picture of the selected acceleration time history after manipulation is given in
Figure 1 while the resulting Fourier spectrum is shown
in Figure 2.
In all the analyses the seismic signal is assumed
to be applied at the outcrop of the deposit. The corresponding bedrock motion was calculated by means of a
deconvolution analysis performed by the code EERA,
as described in the next paragraph.
A 60-m thick ideal deposit of soft clay is assumed
as the reference soil profile, characterised by the following physical and mechanical parameters: plasticity
index IP = 44%, unit weight of volume = 17 kN/m3 ,
overconsolidation ratio in terms of mean effective
stress R = 1.5, small strain shear stiffness GO =
variable with depth, Poissons ratio = 0.25, damping ratio D = variable with depth, coefficient at rest
K0 = 0.6. The water table is assumed at the ground

surface.
A circular tunnel, located at a depth of 15 m and
characterised by a 10.10 m diameter is selected as
the reference underground structure for the present
study. The lining is assumed to be composed by 0.50 m
thick precast concrete segments characterised by the
following parameters: Youngs modulus El = 38 GPa,
Poissons ratio l = 0.25, damping ratio Dl = 5%.
3
UNCOUPLED APPROACH
The 1D ground response analysis was performed

with the code EERA (Bardet et al. 2000) that analyses
the vertical propagation of shear waves in a onedimensional layered system based on an equivalentlinear visco-elastic scheme. It assumes that the shear
modulus G and damping ratio D are function of shear
strain amplitude .
The adopted profile of the small-strain shear stiffness GO with depth (Figure 3) was evaluated by the
relationship proposed by Viggiani (1992) as a function
of the in situ mean stress, R and IP.
Figure 4 shows the curves of the variation of the normalised shear stiffness and damping ratio with shear
strain , defined according to Vucetic & Dobry (1991).
A total number of 31 layers was adopted to discretise
the soil stratum.
Figure 5 shows the results of the analysis in terms of
maximum shear strain max , normalised shear stiffness
G/GO , damping ratio D and maximum acceleration
amax with depth.
Values of max and G obtained at the depth of
15 m, i.e. at the tunnel depth, were subsequently used
Figure 1. Modified acceleration time history scaled at

0.35 g.
Figure 2. Frequency-filtered Fourier spectrum.
Figure 3. Profile of the small-strain shear stiffness GO .
616
to evaluate the increment of hoop forces and bending moments acting on the tunnel lining during the
earthquake, according to selected analytical solutions
proposed in the literature.
Here, the solutions proposed by Wang (1993) to
predict the response of the structure are taken into
consideration for both full-slip and no-slip conditions.
The maximum increment of hoop force and bending
moment in the transverse direction of the tunnel are
given by:
mobilised soil Youngs modulus (evaluated with reference to the previously calculated mobilised shear
modulus G) and the Poissons ratio (assumed equal
to 0.5) in undrained conditions, respectively. r is the
tunnel radius. K1 and K2 are given by the following
expressions:
for full-slip conditions and by:

where:
for no-slip conditions, the bending moment being
the same for the two cases. Eu and u indicate the
are the compressibility and flexibility ratios and t and

I the thickness and the moment of inertia of the tunnel
lining, respectively.
Table 1 summarises the computed increments of
hoop force and bending moment for both full-slip and
no-slip conditions.
4
Figure 4. Modulus reduction curve G/GO and variation of
damping ratio D with shear strain .
COUPLED APPROACH
The coupled analyses were performed by the Finite

Element commercial code PLAXIS, which describes
Figure 5. Results of the 1D ground response analysis performed with EERA.
617
Table 1. Increments of hoop force and bending moment

according to Wang (1993).
Full-slip conditions
No-slip conditions
Nmax
(kN/m)
Mmax
(kNm/m)
Nmax
(kN/m)
Mmax
(kNm/m)
159
802
473
802
Figure 7. Fourier spectra computer by EERA at different

depths and high-energy frequency interval.
Figure 6. Calibration of the G and D profiles assumed in

the FEM analyses on the basis of EERA results.
the mechanical behaviour of the soil in terms of

effective stress.
In order to perform a comparative analysis with the
EERA results a linear visco-elastic constitutive model
was selected. In the constitutive model viscous damping was accounted for by the Rayleigh formulation.
4.1
Calibration of the visco-elastic parameters
The constitutive model employed in the FEM analyses

makes use of constant-value elastic and viscous parameters for each sub-layer of the discretised deposit. In
this respect, it is of paramount importance to define
appropriate values of these parameters as a function of
the strain level attained in the soil deposit during the
earthquake.
In this paper a recently developed calibration procedure of the visco-elastic parameters to be assumed
in dynamic FEM analyses is adopted (Amorosi et al.
2007). G and D profiles are set in such a way to match
the corresponding profiles obtained by the free-field
EERA analysis. To this end the numerical model is
divided into a relatively large number of sub-layers in
order to obtain an as close as possible correspondence
(Figure 6).
Figure 8. R e R profiles assumed in the FEM analyses on

the basis of EERA results.
The profiles of the Rayleigh coefficients R e R are

obtained correspondingly, according to the following
relationship with the damping ratio D (e.g., Clough &
Penzien 2003):
where n e m are the circular frequencies related to

the cyclic frequency interval fn fm . The frequency
range was determined by selecting the interval where
the Fourier spectra computed by EERA at different
depths were characterised by the highest energy content (indicated in the box of Figure 7). The resulting
R e R profiles are shown in Figure 8.
618
Figure 9. Mesh employed in the FEM analyses.
Figure 10. Detail of the mesh around the tunnel section.
4.2
Numerical model
The dimension and boundary conditions of the 2D

FEM numerical model were set up after a considerable number of preliminary analyses performed to
minimise the influence of boundary conditions on the
computed results.
The mesh employed in the present study is reported
in Figure 9: it is characterised by a width equal to 8
times its height. The base is assumed to be rigid and
at the lateral sides the viscous boundaries proposed by
Lysmer & Kuhlmeyer (1969) were used.
The domain was discretised in 2431 15-node plane
strain triangular elements. In the central part of the
mesh, where the tunnel is located, the element dimension h always satisfies the condition:
Figure 11. Comparison between EERA and PLAXIS
free-field soil response analyses at 15 m depth.
where VS is the shear wave velocity and fmax is the

maximum frequency of the seismic signal.
A detail of the mesh around the tunnel is shown in
Figure 10.
The following stages were simulated in the numerical analysis:
first, the tunnel was excavated in undrained conditions imposing a volume loss of 0.4%;
next, the lining was installed and the post-excavation
consolidation phase was studied;
the select seismic signal was then applied at the
bottom of the model in undrained conditions.
In all the static stages of the analysis the soil stiffness for each sub-layer was selected scaling down the
value of the very small strain shear stiffness GO by a
factor of 0.45 to account for an average shear strain
level involved in the excavation stages of = 0.1%.
An interface was activated between the soil and the
lining, characterised by a normal and tangent stiffness
corresponding to that of the adjacent soil: such a condition can be considered similar to the so-called no-slip
conditions of the Wangs solutions.
4.3 Free-field soil response results

In this paragraph a preliminary comparison between
free-field soil response results at the tunnel depth is
provided to check the consistency between the 1D and
2D proposed approaches. In this case the 2D model
does not incorporate the tunnel and, as such, the result
can be directly compared to that of the corresponding
1D free-field analysis.
Figure 11 shows the acceleration time histories and
the Fourier spectra computed by EERA and PLAXIS.
A satisfactory agreement is obtained between the
two solutions, demonstrating the effectiveness of the
proposed calibration strategy.
619
Figure 12. Increments of hoop force and bending moment

during the seismic event.
4.4
Results of the coupled analyses
The maximum increment of the seismic-induced absolute hoop forces and positive and negative bending
moments in the tunnel lining were computed by the
2D FEM coupled analyses and are represented in
Figure 12. The results are reported as a function of
the angle indicated in Figure 12 and defined positive
in counter-wise direction.
Results indicate a satisfactory match between the
maximum increments of hoop force and bending
moments predicted by the visco-elastic FEM solution (Nmax = 423 kN/m e Mmax = 724 kNm/m)
and those resulting from the Wangs solutions for the
no-slip case.
5
CONCLUSIONS
In this paper the transverse dynamic response of a shallow tunnel subjected to seismic actions is investigated
by means of uncoupled and coupled approaches.
The former approach, traditionally used in the engineering practice, combines a free-field site response
analysis with a set of analytical solutions to evaluate
the maximum increment of hoop force and bending
moment in the tunnel lining. This approach is based
on a limited numbers of parameters to describe the
earthquake, the soil behaviour and thawt of the tunnel
lining. The reliability of the uncoupled approach is
supported by the satisfactory performance of a number

of underground structures designed based on it.
Nowadays the availability of relatively sophisticated commercial FEM codes and the continuous
increase in computational power allows to approach
the same problem by means of a single analysis which,
at the same time, accounts for the seismic site response,
the soil-structure interaction and the actions in the tunnel lining. This coupled approach is characterised by
a far more detailed description of the seismic actions
at the tunnel depth as compared to the uncoupled one
and allows to describe the soil behaviour by means of
effective stress based constitutive models. Concerning this last feature in this paper a simple assumption
was made: linear visco-elasticity was assumed for
the clayey deposit under study, where viscous damping was accounted for by the Rayleigh formulation.
Although very simple this constitutive assumption
already poses some problems in the calibration of
the stiffness and Rayleigh damping parameters. In
fact, it is a well established fact that the profiles of
these parameters with depth crucially influence the
results of any FEM dynamic analysis (e.g.: Woodward & Griffiths 1996). Related to this, a strategy for
the calibration of the stiffness and Rayleigh damping
parameters is proposed in the paper. The results of the
coupled analysis based on this calibration satisfactory
match those obtained by the uncoupled approach in
terms of hoop force and bending moments in the lining.
In the Authors opinion this encouraging result represents a necessary condition to extend the use of
the proposed calibration strategy to the case of more
complex constitutive assumptions including plasticity.
REFERENCES
Amorosi, A., Boldini, D., Elia, G., Lollino P. & Sasso M.
2007. Sullanalisi della risposta sismica locale mediante
codici di calcolo numerici. IARG 2007, Salerno.
Bardet, J.P., Ichii, K. & Lin C.H. 2000. EERA-A computer
program for Equivalent-linear Earthquake site Response
Analyses of layered soils deposits. User Manual.
Clough, R. & Penzien, J. 2003. Dynamics of Structures.
Computers and Structures Inc.
Lysmer, J.& Kuhlemeyer, R.L. 1969. Finite dynamic model
for infinite media. ASCE EM 90: 859877.
PLAXIS. 2003. Reference Manual, version 8.
Viggiani, G.M.B. 1992. Small strain stiffness of fine grained
soils. PhD thesis. City University. London.
Vucetic, M. & Dobry, R. 1991. Effects of the soil plasticity
on cyclic response. Journal of Geotech. Eng. Div., ASCE
117(1): 89107.
Wang, J.N. 1993. Seismic design of tunnels: a state-of-the-art
approach. Monograph 7, Parsons, Brinckerhoff, Quade &
Diuglas Inc., New York.
Woodward, P.K. & Griffiths, D.V. 1996. Influence of viscous
damping in the dynamic analysis of an earth dam using
simple constitutive models. Computers and Geotechnics
19(3): 245263.
620
Investigating the influence of tunnel volume loss on piles using

photoelastic techniques
W. Broere
A. Broere BV, Amsterdam, The Netherlands
J. Dijkstra
ABSTRACT: This paper presents the results of plane strain model tests on tunnelpile interaction in a photoelastic material. The effects of volume loss are simulated by contracting a tunnel. The soil in this test is
represented by crushed glass. This allows for the determination of stresses in the model by the photoelastic
method. The influence of the stress change in the soil due to volume loss is shown, as well as the effect on three
rows of piles at varying distance from the tunnel. From the tests it is clear that significant stress changes occur
close to the pile tips.
INTRODUCTION
The upcoming boring of the North-South metro line in

Amsterdam (Netzel & Kaalberg 2000, Kaalberg et al.
2005) will involve TBM excavation close to the pile
tips of the wooden piles on which much of the historic inner city is founded. The influence of a volume
loss due to tunnelling on the bearing capacity of these
piles is relatively uncertain and hard to quantify even
by numerical means (Broere & van Tol 2006). This
prompted a different approach, to visualise the influence of volume loss on the stresses around the pile
tips using the photo-elastic method. Photoelasticity
has been used extensively to quantify stresses in homogeneous materials (e.g. Frocht 1948). And is extended
to particle assemblies in index matching liquids by
Drescher (1976).
A small research effort was started in 2002 to
attempt to quantify the influence of a volume loss
on the stresses close to pile tips in a photoelastic
scale model. As the interpretation of the measurements
proved more complicated than originally anticipated,
it took considerably more effort to finalise the analysis and derive the complete stress state. An overview
of these complexities is given in Petrucci & Restivo
(2007). Standing & Leung (2005) had similar considerations as outlined above when they did comparable
photoelastic scale model tests, but they decided to
invert the problem and only qualify the influence of
pile installation on the stress in existing tunnel linings.
STRESS MEASUREMENTS
It is possible to obtain detailed information on the

stress distribution of a granular material using the photoelastic measuring method. The soil is replaced by
grains of a photoelastic material, in this case glass particles (Wakabayashi 1957, Drescher 1976). Crushed
glass behaves similar to sand particles, although, the
grains are more angular. Therefore it is a reasonable substitute to investigate sand behaviour. However, other authors, most noticably Lesniewska &
Sklodowski (2005) argue that when uniform glass
beads are used instead of crushed glass the stress trajectories a long failure planes can be better visualized.
When compared to the photoelastic measurement
methods for continuous materials the quantification
of the stress paths for a granular material is hampered.
First, in order to eliminate light scatter the pores have
to be filled with a liquid which has a similar refraction
index as the glass particles. Secondly the analysis of
the fringe patterns is impossible because of the stacked
nature of the sample. Each layer of grains produces
its own fringe pattern resulting in a mix of fringes
which cannot be interpreted. The first condition leads
to the qualitative visualization of these stress paths by a
circular polariscope, see i.e. Wakabayashi (1957). For
the second problem, quantification of the stress paths,
Allersma (1982) developed an automated polariscope.
In general, granular material does not behave
elastically, as is often assumed when formulating a
621
Figure 1. Front view of model setup, all measures in mm.
stressstrain relationship. Therefore it is not possible

to derive the stress increments from strain measurements. The stresses have to be measured independently
of the strains.
The method developed by Allersma (1987) and
Allersma & Broere (2002) for measurement of the
stress in a photoelastic material is followed. The
method already published by Allersma in 1982, is
most comparable to phase-stepping photoelasticity as
described in e.g. Ajovalasit et al. (1998). The direction of the principal stress and the principal stress
difference along the light path are measured with an
automated circular polariscope. Therefore averaging
of the stress in the direction normal to the plane is
taken into account.
In a photoelastic material the refraction in each
direction is dependent on the normal stress in that
direction. The stress difference between the principal
stress directions, i.e. twice the maximum shear stress,
2 * max , leads to a relative light velocity difference and
subsequently a change in polarisation. This change in
retardation can be measured to obtain the shear stress
in a point.
The maximum shear stress in a point can be derived
from the elliptical polarisation of the light. The light
changes from circular polarised light into elliptical
polarised light when travelling through the sample,
caused by the stress in the material.
If at one or more points the complete stress state is

known, the absolute values of the stress tensor can be
determined by integrating the equilibrium equations.
Unfortunate, this scheme is very sensitive for experimental noise, which hampers the results. In the present
tests the vertical stress is calculated from the measured
surcharge at the top of the sample.
3 TEST SETUP AND TEST PROCEDURE
3.1 Test setup
The test setup consists of a model container with inner
dimensions of height width of 690 mm 530 mm
and a depth of 70 mm. In Figure 1 a front view of the
test setup is given. The front and rear sides are made of
glass. The rectangular piles are made of stainless steel
with a cross sectional area of 10 10 mm. Two rows,
with a c.t.c. distance of about 25 mm, of three piles
are installed at respectively 0.5Dtunnel , 1Dtunnel and
1.5Dtunnel , simulating a piled foundation. Also a tunnel is embedded at 345 mm from the side and 232 mm
from the bottom of the container.
The tunnel consists of an upper and a lower part
milled of stainless steel which can be pushed out vertically by an inner stepper motor. Therefore, the volume
change of the tunnel is primarely caused by vertical
contraction. The mechanical realisation of this tunnel
622
Figure 2. Mechanical realisation of the tunnel.
can be seen in Figure 2. An additional surcharge is

applied on a steel beam to increase the stress level in
the sample. This surcharge is applied by spring loading the beam, i.e. between a fixed boundary and the
beam a mechancial spring is applied, such that small
displacements in the sample do not totally unload the
boundary. Between the spring and the fixed boundary
a load cell is monitoring the applied load.
The box is filled with crushed glass particles and
the index matching liquid. The particle size of the glass
particles is between 2 and 3 mm. The material behaves
similar to sand of the same particle size as shown
in triaxial tests performed by Allersma. According to
Allersma, the angle of internal friction at constant volume, cv , for angular silica sand of this particle size is
ca. 33 and for crushed glass ca. 39 .
Figure 3. Principal stress direction and magnitude before

tunnel contraction.
without altering the surcharge, the tunnel is contracted

by decreasing the distance between the upper and lower
tunnel parts by 2 mm or 1.2%. This results in a volume
loss of only 0.6%, as the tunnel contracts vertically
only and not uniformly. This will result in an unloading of the soil around the tunnel. Again a photoelastic
measurement is made.
The mechanical polariscope is mounted on a computer controlled x-y scanner and scans approximately
1200 stress points spaced at 10 mm interval in an area
of ca. 340 mm 340 mm covering half of the tunnel
and the lower part of the embedded piles. The measurement zone is also shown in Fig. 1 (dotted lines).
3.2 Test procedure

Medium dense conditions are obtained by first pouring
the glass into the strongbox, followed by densification
of the sample, and finally pumping the liquid through
the sample from bottom to top. The tunnel is undisturbedly embedded during preparation, while keeping
the diameter at its maximum position. The piles are
pushed in afterwards before applying the surcharge.
A surcharge of 75 kPa is applied at the steel beam after
preparation of the sample is finished. Before starting
the test first an equilibrium stage has to be reached
in which no additional creep is apparent. During this
phase the surcharge is kept constant at 75 kPa.
The objective of this test is to investigate the effect
of volume loss on existing pile foundations, therefore at the beginning of the test the tunnel is in its
opened position. Also the piles are already embedded.
Subsequently a photoelastic measurement is made by
means of a mechanical polariscope. In the next phase,
4 TEST RESULTS
The results of the performed tests are summarized in
three figures. The first Figure, Fig. 3, is showing
the principal stress directions and the magnitude of
the principal stresses measured before tunnel contraction, the second Figure, Fig. 4, is showing the
same data, but now for the situation after tunnel contraction. The principal stress crosses are all plotted
with the same scale, i.e. the length determines the
magnitude. Finally, the last figure, Fig. 5, plots the
contour plot of the difference in the measured maximum shear stress. The before measurement is taken
as reference, resulting in a positive value for the difference when the maximum shear stress in a point
before contraction was higher than after contraction
and logically a negative value when the value afterwards is higher than before. The measured surcharge
623
Figure 4. Principal stress direction and magnitude after

tunnel contraction.
points are horizontally and vertically spaced at 10 mm.

Firstly the principal stress rotations right on top and
below the tunnel are all negligible, the tunnel is loading
the soil in compression, while to the side large stress
rotations occur. The behaviour is therefore ver-tical
symmetrical.
Another observation is that the spacing between the
piles is such that arching seem to occur, spots without any rotation and with a lower stress magnitude
can be observed. Clearly, a spacing of about seven
times the pile diameter is still sufficient to cause this
effect, however scale effects cannot be excluded as the
grains are relatively large compared, dgrains = 23 mm,
to the pile diameter which is 10 mm. The most right
pair of piles (remember two rows of piles are intstalled
behind eachother when seen from the front side), or the
pile row next to the boundary of the model container
is distributing its load immediately in the side wall.
As directly below a large undisturbed zone is seen.
This is clearly a boundary effect. Lastly the pile closest to the tunnel is influenced by the tunnel, as between
the pile and the tunnel a direct path of parallel stress
directions can be seen. This pile row is at a tiptunnel
distance of about 12 cm or 12Dpile or 0.75Dtunnel .
4.2 After tunnel contraction
Figure 5. Maximum shear stress difference (kPa) beforeafter.
after the first measurement is 74.8 kPa, while after the

second measurement the reading for the surcharge is
74.3 kPa. Therefore, the changes found between the
two measurements are mainly caused by the tunnel
contraction and not by additional creep in the sample. As already elaborated, the angle of the principal
stress and the maximum shear stress are direct measurements, while the principal stress magnitude is
processed from the stress integration algorithm. All
geometries, like piles and tunnels shown in the figures
are plotted to scale at the proper position.
4.1
Before tunnel contraction
When studying the results of the measurements taken

before tunnel extraction, given in Figure 3 several
things are noticed, for sense of scale the measurment
Again the principal stress rotation and magnitude are

plotted in Figure 4, this time for the situation after
tunnel contraction. All former observations still apply,
pile arching seems to still occuring, the closest pile row
is still influenced by the tunnel and is unexpectedly
in a higher stress regime. However, closer inspection
reveals that the band of stress between the closest pile
row and the tunnel seems smaller in width i.e. the stress
intensity is lower. Also, the stresses directly below the
tunnel seem changed also, the undisturbed zone below
is smaller. From an experimental point of view this
can be expected, because as the tunnel is decreased in
height, the largest changes are expected in the vertical
stress distribution. Not only unloading on the top side
of the tunnel occurs but also an elastic rebound on the
bottom.
4.3
Shear stress difference
It is rather difficult so observe differences in magnitude in principal stress from the already shown
figures, therefore one additional figure is presented. In
Figure 5 the difference in measured maximum shear
stress between the situation before and after tunnel
contraction is plotted as the difference is more clear
from this data than from the processed data. The numbers plotted next to the x-axis and y-axis are the
distance from the origin of the measurement in mm.
The values given in the colourbar of the contourplot are in kPa, positive values indicate a lower stress
than before tunnel contraction, negative values point
towards a higher stress than before contraction.
624
It is immediately clear from this data that an

increase in stress around the pile tips of the three pile
rows is found. Another observation is that the first two
pile rows are most effected by the contraction. The
zones with the highest stress concentrations, in this
case around the pile tips and at the point at which the
stress arch reaches the tunnel are taking all the stress
change, while the general field does hardly change at
all. The last finding further supports the observation
that creep in the sample during the tests is kept at a
minimum. The increase in maximum shear stress can
be explained by the fact that in the model setup due
to volume loss the soil immediately begins to create
movements of soil below the pile rows resulting in
negative friction on the piles.
5
DISCUSSION
It is our opinion that for the case considered, a scaled

plane strain situation with rather large grain size and
low stress situation, the results are still convincing.
The effects of nearby tunnelling are seen in the measured stress field. Increase of maximum shear stress
can be explained by negative friction on the piles. However, the change of the horizontal or vertical stress
is more difficult to acquire, both due to difficulties
from noise in experimental data, as measuring properties of light is always very sensitive to experimental
errors, and due to the integration of the measured stress
field. Therefore, no conclusive answer can be given at
this stage.
All pile rows in the presented setup suffer from the
tunnel contraction effects, but the first row is clearly
most influenced, not only by the tunneling but also in
the reference situation interaction is seen between the
two structural elements. The influence of the tunneling
on the last pile row is minor, but this pile row is already
located at 1.5Dtunnel to the center line of the tunnel.
When compared to the results of Standing & Leung
(2005) a bit more information is derived from the
measurements. In the current setup the model scale,
especially when considering the dimensions of the
pile, the model container and tunnel, are more realistic.
Similar to Standing & Leung (2005) in our measurement nearby piles attract more stress lines, which
reduce in intensity after the tunnel contraction. This
indicates a stress release near the pile, which is consistent with the increase in shear along the pile as the
pile starts to settle.
6
CONCLUSIONS
The influence of a contracting tunnel on the stresses

near the tips of displacement piles has been studied
using photoelastic model tests. From these tests, the
changes in principal stresses and shear stresses near
the tunnel and piles have been quantified. Given the
limited size of the scale model, and the relatively large

grains used to improve the visibility of the photoelastic
effect, some limitations of the model have come to light
and these have been discussed. Especially as there is
some evidence of stress arches between the different
piles, the results of these tests cannot be translated
directly to a field situation but should be interpreted
with care.
Nonetheless, the tests show a clear influence of
a volume loss of the tunnel of 0.6% on the stresses
near the pile tips up to 1Dtunnel . Here stress bands
are decreasing in width, indicating a lowering of the
principal stresses, and at the same time an increase
of the shear stresses along the pile shafts is seen for
these piles. These observation corroborate those of
Standing & Leung (2005), but also indicate that for
displacement piles close to failure the influence zone
of a tunnel volume loss is wider than suggested there.
Contrary to their findings not only piles with their toes
inside the influence zone defined by Jacobsz et al.
(2003) are affected, but also those just outside of that
zone.
This suggests that the influence zone to be taken into
account for displacement piles, which depend on both
end bearing and skin friction, might be slightly larger
than for bored piles, which are mostly founded on end
bearing alone. To determine this would require a combination of further field observations, model testing
and numerical modelling of the problem.
REFERENCES
Ajovalasit, A., Barone, S., & Petrucci, G. 1998. A review
of automated methods for the collection and analysis of
photoelastic data. Journal of Strain Analysis 33(2), 7591.
Allersma, H. 1982. Determination of the stress distribution in assemblies of photoelastic particles. Experimental
Mechanics 22(9), 336341.
Allersma, H. 1987. Optical analysis of stress and strain
in photoelastic particle assemblies. Ph. D. thesis, Delft
University of Technology.
Allersma, H. & Broere, W. 2002. Optical analysis of stress
around a penetrating probe in granular material. In Physical Modelling in Geotechnics: Proceedings of the international Conference on Physical Modelling in Geotechnics,
pp. 149154.
Broere, W. & van Tol, A. 2006. Modelling the bearing
capacity of displacement piles in sand. Proceedings of
the Institution of Civil Engineers, Geotechnical Engineering 159(3), 195206.
Drescher, A. 1976. An experimental investigation of flow
rules for granular materials using optically sensitive glass
particles. Gotechnique 26(4), 591601.
Frocht, M. 1941 and 1948. Photoelasticity volumes 1 and 2.
New York: John Wiley & Sons.
Jacobsz, S., Standing, J., Mair, R., Soga, K., Hagiwara, T., &
Sugiyama, T. 2003. Tunnelling effects on driven
piles. proc. int. conf. on response of buildings to
excavation-induced ground movements. In F. Jardine
625
(Ed.), CIRIA Special Publn 201, pp. 337348. CIRIA,

London.
Kaalberg, F., Teunissen, E., Van Tol, A., & Bosch, J. 2005.
Dutch research on the impact of shield tunnelling on
pile foundations. In Proceedings of the 5th international
Construction in Soft Ground, pp. 123131.
Lesniewska, D. & Sklodowski, M. 2005. Photoelastic investigation of localization phenomena in granular materials.
In Powders and Grains, pp. 6972.
Netzel, H. & Kaalberg, F. 2000. Numerical damage risk
assessment studies on masonry structures due to TBMtunneling in Amsterdam. In O. Kusakabe, K. Fujita,
and Y. Miyazaki (Eds.), Geotechnical Aspects of Under-
ground Construction in Soft Ground. Balkema. preprints

pp. 235244.
Petrucci, G. & Restivo, G. 2007. Automated stress separation along stress trajectories. Experimental Mechanics
published online.
Standing, J. & Leung, W. 2005. Investigating stresses around
tunnels and piles using photo-elasticity techniques. In
Proceedings of the 5th international Symposium on
Soft Ground, pp. 171177.
Wakabayashi, T. 1957. Photoelastic method for determining
of stress in powdered mass. In Proceedings of the seventh
Japanese National Conference on Applied Mechanics,
pp. 153158.
626
Assessment of tunnel stability in layered ground

P. Caporaletti
Atkins Geotechnics, London, UK (formerly University of Rome La Sapienza, Rome, Italy)
A. Burghignoli
University of Rome La Sapienza, Rome, Italy
G. Scarpelli
Universit Politecnica delle Marche, Ancona, Italy
R.N. Taylor
City University, London, UK
ABSTRACT: The stability of a circular tunnel in layered ground, with both fine-grained and coarse grained
soils below the water table, is investigated experimentally and theoretically. Centrifuge tests have been carried
out at City University, London, investigating in detail tunnelling effects in layered ground, in terms of both soil
displacements and strains and in terms of the kinematics at failure. Also, analytical upper bound solutions for
layered ground which closely reflect the observed failure mechanism of the tunnel have been derived independently. The results from all different approaches have been compared by emphasising the effects of the different
hypotheses on the assessment of tunnel stability.
INTRODUCTION
The assessment of tunnel stability is an important

issue considering the catastrophic effects induced by
the tunnel collapse, especially when it concerns urban
areas. Surface surcharge loadings may occur in practice where tunnels are excavated below pre-existing
structures through fine-grained soil which is overlain
by granular material: typical conditions of excavation
in urban areas. Although the weights of the overlying
coarse-grained materials are easily taken into account
when assessing the tunnel stability, the contribution of
their strength and stiffness on tunnel stability is often
ignored or too difficult to be quantified.
The problem will be approached through the upper
bound plasticity theory together with results of centrifuge tests on model tunnels. The case of a tunnel
excavated within an over consolidated clay deposit
overlain by a sandy layer is considered.Tests included a
masonry wall setting in the coarse grained upper layer
of soil and thus also gave additional information on the
interaction of the tunnel with pre-existing buildings.
This study began from literature references, following two different approaches. In the first case the sandy
layer has been taken into consideration as surcharge
acting on the top of the clay layer where the tunnel
is excavated. In the second case, the sandy layer is
properly considered by assigning its appropriate unit

weight, thickness, and strength in order to simulate
the real initial stress state of soil around the cavity and
to emphasise the different behaviour in terms of both
undrained and drained conditions.
2
STABILITY OF A SHALLOW TUNNEL
The excavation of a shallow tunnel is clearly a threedimensional problem. Neglecting the effect due to
the volume loss at the front face of excavation, this
problem might be analyzed under plane conditions.
Figure 1 shows an idealization of shield tunnelling,
where a circular tunnel of diameter D is constructed
with a depth of cover C. The tunnel lining is regarded
Figure 1. An idealization of shield tunnelling.
627
as rigid and in front of it the tunnel heading is represented by a cylindrical cavity of length P in which a
uniform fluid pressure t acts, and a uniform pressure
s acts on the soil surface (Figure 1a). For relatively
big value of the distance from the tunnel lining to the
tunnel face, the problem can be idealized under plane
conditions (Figure 1b).
The collapse of tunnel heading is usually a sudden event caused, for example, by a sudden loss of
tunnel support pressure. Then, the stability of tunnels
in fine-grained soils can be evaluated by referring to
undrained conditions, while the stability of tunnels in
coarse-grained soils can be analyzed under drained
conditions.
the upper and lower bounds of N do not change significantly with D/Su . Below C/D equal to 3, there is
a larger spread but the lower bound for D/Su = 0 can
be adopted as a safe criterion to calculate the required
tunnel support pressure: see Figure 4, by referring to
the upper and lower bound curves after Davis et al.
(1980). If the ratio D/Su is sufficiently large then the
collapse will take place for any value of uniform tunnel pressure. Sloan & Assadi (1992) also concluded
that for tunnels with C/D>3, these solutions are not
fully reliable since the stability of deep tunnels is usually related to a complicated local collapse, involving
both elastic and plastic deformation, with only small
settlements taking place at the ground surface.
2.1
2.2 Stability of a tunnel in drained conditions
Stability of a tunnel in undrained conditions
The stability of a tunnel excavated in undrained soils

may be assessed using the stability ratio, N:
where is the soil unit weight; H is the tunnel axis

depth from the ground surface; s and t are respectively the surface surcharge pressure and the tunnel
support pressure (according to Figure 1); and Su is the
soil undrained shear strength at the tunnel axis level.
Davis et al. (1980) estimated the safety of a shallow
tunnel excavated without internal support by considering three different shapes of shallow underground
opening, for which upper and lower bound stability solutions were derived. The soil strength Su was
assumed to be constant with depth. Sloan & Assadi
(1992) presented a rigorous study taking into account
for the variation of the undrained shear strength
with depth. The theoretical approach of upper and
lower bound solutions were compared to experimental
results from a comprehensive study in clays conducted
at Cambridge University over the last decades e.g.
Kimura & Mair (1981) who gave a range of design
lines to estimate the stability ratio for different tunnel
geometries, mainly focusing on undrained conditions.
Generally speaking, the bounding solutions gave
good estimates of the collapse load and supported the
use of classical limit analyses for undrained conditions. It should be noted that the stability of a plane
strain unlined circular tunnel (LR /D) is more critical than that for the lined circular tunnel heading with
LR /D = 0 and the analysis is therefore conservative.
Previous literature proposed different charts where
the tunnel pressure at collapse is related to soil
mechanical properties and tunnel geometry: the term
(s t )/Su or the stability ratio N is given as a function
of C/D for different ratios of D/Su <4. If Su is constant with depth, then for values of C/D greater than 3
Atkinson and Potts (1977) derived kinematic upper

bound, and statically admissible lower bound plasticity solutions for the two-dimensional idealization in
Figure 1. Only dry sand was considered, but the theoretical solutions, based on effective stress, may be
applied to saturated sands provided that the pore water
is stationary and the pore water pressures around the
tunnel are known. The collapse of a tunnel in saturated
sand is then simply the sum of the pressures predicted
by either the upper or lower bound solutions and the
pore water pressure. The situation is rather more complicated if there is a steady state, or transient, seepage.
The authors gave equations to obtain the dimensionless
ratio t /D as function of C/D: the upper bound solution (being inherently unsafe) gives lower value than
the lower bound solution which is inherently safe. The
authors concluded that the support pressure is independent of the ratio C/D. They also obtained good
comparisons between these theoretical solutions and
experimental data from centrifuge tests.
2.3 Stability of a tunnel in layered ground

As previously discussed, as far as the authors are aware
there is currently no standard procedure to account
for the contribution to the tunnel stability of overlying
layers of different materials.
Grant & Taylor (2000) studied the stability of tunnels excavated in clay with overlying layers of coarse
grained sands and gravels, referring to the design line
for clay only given by Kimura & Mair (1981) for
LR /D. Experimental data from centrifuge tests
with an upper layer of loose sand fitted the design line
for clay only, indicating that it may act as surcharge
loading and not contribute to the stability of tunnel
except in terms of weight. In contrast, the presence
of a significant thick layer (at least 1D) of relatively
dense coarse grained material, combined with little
cover above the tunnel crown (C<2D), will aid the
628
Figure 2. Picture of the experimental model: wall-ground

interaction configuration.
stability of a tunnel, and the reasonable value of N

equal to 4 might be assumed.
3
CENTRIFUGE MODEL TESTS
Centrifuge model testing has proved a very useful tool

to examine the behaviour of shallow tunnels in layered
ground, in terms of soil displacements and strains, such
as in terms of kinematic failure. A number of tests have
been performed at City University, by modelling both
the greenfield conditions and the interaction problem
between tunnels and pre-existing masonry structures.
Details of the apparatus and procedures of tests have
been given elsewhere (Caporaletti 2005; Caporaletti
et al. 2006) and only the essentials features will be
described in this paper.
All centrifuge tests were performed under plane
strain conditions: Figure 2 shows a schematic of the
centrifuge model. A layer of pre-consolidated kaolin
is overlain by a layer of medium dense sand. The
kaolin slurry was pre-consolidated by applying a onedimensional load in a press to a vertical effective stress
of 500 kPa, before allowing the clay to swell back to
250 kPa. The overconsolidation ratio ranged between
1.4 to 2.8 with depth. The stratum of medium dense
sand was made by manual pluviation and then frozen.
A 50 mm cavity was cut through the clay layer using a
thin walled-cutter and lined with a thin rubber membrane of negligible stiffness and strength.All tests were
carried out at the same scale factor Ng =160 in order to
model a real tunnel of diameter equal to 8 m, excavated
in a clay layer of 22 m depth overlain by a sand layer
of 12.8 m depth, with the tunnel axis at about 23 m
from the ground surface. The effects of tunnelling on
pre-existing structures were studied by modelling a
Figure 3. Front view of the model during the test from image
processing: a) before reduction of t ; t0
= 386 kPa; b) at
tunnel collapse with t
= 25 kPa.
completely buried thin masonry wall perpendicular to

the tunnel axis, with foundations just at the sand-clay
interface. It represents a stiffer and heavier inclusion
in the sand layer.
During the centrifuge spin-up the tunnel air pressure
t applied a uniform radial total stress into the cavity
to balance the total stress at tunnel axis level:
where S is the unit weight of sand; C is the unit weight

of clay; Cs is the thickness of the sand layer; Cc is the
cover above the tunnel in the clay layer; and R is the
radius of tunnel as shown in Figure 2. This tunnel pressure represents the compressed air, bentonite slurry or
a shield used in practice to achieve the tunnel stability
during the excavation process. Water was supplied to
the model through a stand-pipe to maintain a predetermined water-table throughout the model: the water
level was set up at different depths from the ground
surface in order to evaluate the influence of the different effective stress distribution on the tunnel stability.
Equilibrium pore pressure was measured by miniature
pore pressure transducers around the cavity. The tunnel excavation has been performed by reducing the air
tunnel pressure until the cavity collapse, in a period of
34 minutes at a rate of approximately 85 kPa/min.
Digital images were taken every second by using a
video camera fixed on the swinging platform to view
the front face of the strongbox during tests and to follow the movements of black targets pushed both in
clay and in sand. Figure 3 shows images of the model
in flight.
In spite of the relatively high value of the kaolin
permeability, the stratum of clay performs general
629
Figure 4. Stability ratio, N: experimental data against

theoretical solutions for clays under plane conditions.
undrained conditions during the reduction of tunnel

pressure. However, considering that the upper layer
of sand represents a significant boundary of drainage
for the lower kaolin layer, and considering the small
thickness of the clay cover above the tunnel cavity,
local drained conditions in this upper part of the kaolin
layer were likely in those tests.
3.1
Centrifuge test results
The value of the tunnel pressure, t , has been evaluated

for each test, in order to compare experimental results
to the guidelines from literature references.
The cavity was excavated in clay, and so design lines
for homogenous undrained soils have been considered
first. Figure 4 and Figure 5 present the comparison in
terms of the stability ratio, N, and of t /Su depending
on the soil mechanical property and tunnel geometry.
The profile of the undrained shear strength with depth
has been calculated following the equation (Koutsoftas
& Ladd 1985):
were v is the vertical effective stress, and OCR is

the overconsolidation ratio. Considering the relatively
small thickness of the clay cover, a constant value of
Su has been chosen in order to compare test results
with theoretical solutions by Davis et al. (1980) for
the corresponding value of the term c D/Su = 2. In
the interpretations analyzed afterwards, the constant
value of Su refers to a characteristic depth respectively
equal to 5 m and 1.8 m from the sand/layer interface
(Ribacchi et al. 1993).
Experimental data represented by open squares
have been analyzed by simply assuming the sand layer
as surcharge acting on the top of the clay layer, s , and
it is equal to its own weight. Then, the tunnel cover,
C, is only equal to Cc . On the contrary, full symbols
Figure 5. Tunnel pressure, t : experimental data against

theoretical solutions for clays under plane conditions.
represent the experimental results analyzed by properly including the sand layer in the evaluation of the
tunnel cover. The sand layer is allowed to make a contribution to the tunnel stability as a soil with similar
strength to the clay. The surface surcharge pressure,
s , is therefore equal to zero, and the tunnel cover
C = Cs + Cc . The comparison between the two different data interpretations emphasizes that the ratio of
the cover to the diameter of the tunnel is clearly a significant aspect on the assessment of the pressure at
collapse. A good fitting of experimental data to previous plasticity solutions from literature references was
obtained only when the tunnel cover is properly considered (full symbols). While, the open squares are always
out of the range given by theoretical lower and upper
bounds. However, this agreement definitely worsens
if the real value of Su at tunnel axis is assumed instead
of the value at the characteristic depth as previously
explained.
Afterwards, experimental data have been compared
to theoretical solutions for cavity excavated in sands.
Design lines after Atkinson and Potts (1977) are shown
in Figure 6, by distinguish the original lower and upper
bound approaches for dry sands, and the lines obtained
for saturated soils. Once again, the physical results
are clearly out of the two ranges given by theoretical
solutions.
The load factor, which is the ratio between the stability ratio at working conditions and at collapse (the
subscript 0 refers to the beginning of the test before
the reduction of the tunnel support pressure):
has been calculated at different values of the volume

loss, VL , defined as the volume of settlement trough
at the tunnel axis, and plotted in Figure 7.
In order to have a clear representation of experimental data, only results from few tests have been
shown. Test PC2 and test PC8 had greenfield conditions but with two different hydraulic boundaries: the
630
the upper face of kaolin, which did not occur with the
strong wall that lost this contact at value of VL 15%.
In spite of this, the failure mechanism observed in the
soil around the tunnel seemed to be very similar. It is
particularly interesting to emphasise that for all tests
the experimental curves approximately show a horizontal asymptote after volume losses of around 20%.
Therefore, the condition corresponding to the tunnel
collapse has always been taken as that corresponding
to the stage when VL >20%.
3.2 Tunnel mechanism of failure
Figure 6. Tunnel pressure, t : experimental data against
theoretical solutions for sands under plane conditions.
Figure 7. Curves of load factors against volume loss.
water level is very close to the ground surface (PC2),

and very close to the sand/clay interface (PC8). Soilstructure interaction conditions existed in Tests PC6
and PC7, in which were modelled respectively the
weakest and the strongest masonry structure, with
the same hydraulic conditions fixed during test PC8.
As expected, focusing at first on the greenfield tests,
higher values of the tunnel support pressure were measured during test PC2, which means a more unstable
configuration due to the different effective stress distribution experienced as the consequence of differences
in the hydraulic boundary conditions. At the same way,
analyzing the results of soil-structure interaction tests,
the masonry wall built in the sand layer represented
a stiff inclusion within the soil and made the tunnel
less stable. The tunnel collapse occurred for higher
values of the cavity pressure, and the worst condition was reached when the strong wall was modelled.
Looking at the point of intersection between the load
factor curves related to the soil-structure interaction
tests, and the reference greenfield curve, it is evident
that the weakest wall always stayed in contact with
Experimental data have been analyzed in terms of the

effect due to the layered configuration on the field of
soil stress and strain (Caporaletti, 2005; Caporaletti
et al., 2006). The two fine-grained and coarse-grained
materials, sand and clay, have clearly shown different
mechanical behaviours. In particular, the sand tendency to dilate was constrained by the lateral walls
of the strongbox, and resulted in vertical settlement
decreasing with depth. At the base of the sand layer, the
pattern of soil strains indicates expansion near the
tunnel axis. In contrast, vertical settlements within
the clay layer always increased with depth, and nonzero volumetric compressive strains were measured
above the cavity close to the tunnel axis, maybe due to
there not being perfectly undrained conditions locally
maintained during the tests. According to literature
references, the mechanism of failure for tunnel in
clays propagates upwards and outwards from the cavity invert becoming significantly wider than the tunnel
diameter. In sands, failure tends to involve a narrow
chimney, propagating almost vertically from the
cavity up to the ground surface (Mair & Taylor (1997)).
However, in these centrifuge tests the mechanism at
failure for a layered configuration involves a wide area
of soil both in sand and in clay, with pseudo-vertical
settlements at the sand-clay interface. The kinematic
mechanism is clearly characterized by soil displacements pointing towards the cavity on the top of the
sand layer and everywhere within the clay layer, and
by a rigid vertical block translation at the base of the
sand layer, for a thickness of about D/2 (see Figure 8).
The field of ground displacements in sand is strictly
related to soil movements induced at the top of the clay
layer more than to the reduction of the tunnel support
pressure itself.
4 THEORETICAL ANALYSIS
The comparison of experimental data to published literature guidelines has clearly shown discrepancies in
terms of evaluation of the soil behaviour at collapse,
when both fine and coarse grained material strata are
involved in the problem. The effect due to the layered
631
Figure 8. Mechanism of failure from centrifuge tests

(VL
= 20%).
Figure 9. Mechanism for layered ground.
configuration need to be better understood: kinematic

analyses could explain the observed behaviour.
A mechanism related to the problem of stability
of a plane strain unlined circular tunnel excavated in
layered ground has been evaluated. The calculation
relies upon the assumption that failure happens under
drained conditions in sand and undrained conditions in
clay. The soil is idealized as a perfectly plastic material
with an associated flow rule: unit weight, s for sand,
and c for clay, undrained shear strength of clay, Su , and
angle of dilancy of sand equal to the maximum angle of
shear resistance, = . The mechanism is shown in
Figure 9. It starts from that given by Davis et al. (1980)
for clays, but it assigns the appropriate unit weight,
thickness, and strength to the coarse-grained stratum.
Due to the assumption of associated flow, the mechanism shows vertical movements within the sand layer,
according to the experimental data (see Figure 8). Vertical movements are also assumed to be generated in
clay above the cavity, by simplifying the real kinematic observed from centrifuge tests (see Figure 8).
The work calculation is performed by changing the
angle in order to minimize the value of the tunnel
pressure and to achieve an upper bound solution, with
an unsafe estimation of the pressure at collapse.
Figure 10. Theoretical solutions and experimental data.
The theoretical solution for this mechanism is plotted in Figure 10 together with centrifuge results and
literature solutions for homogenous undrained soils:
the stability ratio, N, refers to the actual undrained
strength of clay at the tunnel axis. The cover in sand
has been assumed equal to D/2, following the kinematic mechanism observed during centrifuge tests,
while part of the thickness of the sand layer is evaluated
as an external surcharge. The stabilizing contribution
due to the friction acting in sand is clearly evident,
and its influence on the evaluation of the stability ratio
should not be neglected. Experimental data are clearly
out of the range given by lower and upper bounds for
clays, whereas a satisfactory upper bound is assessed
by adopting the theoretical solution by following the
mechanism herewith proposed.
CONCLUSIONS
Physical modelling and theoretical solutions have been

studied in this research and compared to one another, to
highlight the effect on the tunnel stability due to a layered configuration that involves both fine-grained and
coarse-grained materials, since previous works only
refer to homogenous soils.
In the case of tunnels excavated in clay overlain
by sand, the contribution to the stability due to the
friction acting within the upper layer represents a significant contribution, which should not be neglected.
Coarse-grained material cannot satisfactory be considered as an external surcharge equal to its own weight
only. A significant over-estimate of the tunnel support
pressure to prevent collapse might result if simple reference to the literature theoretical solutions obtained
for homogenous clays is made and the sand layer
treated only as a surcharge. The proposed new mechanism provides a good upper bound to the experimental
data though some simplifications of the proposed kinematic mechanism could be removed, and though it
632
is recognised that a new lower bound is needed to

complete the analysis.
REFERENCES
Atkinson, J.H. & Potts, D.M. 1977. Stability of a shallow
circular tunnel in cohesionless soil. Gotechnique 27(2):
203215.
Caporaletti, P. 2005. Tunnelling in Layered Ground and its
Effects on Pre-existing Masonry Structures. PhD Thesis,
University of Rome La Sapienza, Italy.
Caporaletti, P., Burghignoli, A. & Taylor, R.N. 2006. Centrifuge Study of Tunnel Movements and their Interaction with Structures. Proceedings 5th Int. Conference on
Soft Ground. Amsterdam, The Netherlands: 99105.
Davis, E.H., Gunn, M.J., Mair, R.J. & Seneviratne, H.N. 1980.
The stability of shallow tunnels and underground openings
in cohesive material. Gotechnique 30(4): 397416.
Grant, R.J. & Taylor, R.N. 2000. Stability of tunnels in clay

with overlying layers of coarse grained soil. Proceedings
GeoEng2000. Melbourne, Australia.
Kimura, T. & Mair, R.J: 1981. Centrifugal Testing of Model
Tunnels in Soft Clay. Proceedings 10th International Conference on Soil Mechanics and Foundations Engineering.
Stockholm, Vol. 1: 319322.
Koutsoftas, D.C. & Ladd, C.C. 1985. Design strength of an
offshore clay. Journal of the Geotechnical Engineering
Division, ASCE 111(3): 337355.
Mair, R.J. & Taylor, R.N. 1997. Theme lecture: Bored tunnelling in the urban environment. Proceedings 19th International Conference on Soil Mechanics and Foundations
Engineering. Hamburg: 23532384.
Ribacchi, R. 1993. Recenti orientamenti nella progettazione
statica delle gallerie. XVIII Convegno Nazionale di
Geotecnica. Proceedings 18th National Conference in
Geotechnics. Rimini, pp. 37.
Sloan, S.W. & Assadi, A. 1992. Stability of shallow tunnels in
soft ground. Predictive soil mechanics, Thomas Telford,
London, 1993: 644662.
633
Reinforcing effects of forepoling and facebolts in tunnelling

K. Date
Kajima Technical Research Institute, Kajima Corporation, Tokyo, Japan
R.J. Mair & K. Soga

Engineering Department, University of Cambridge, Cambridge, UK
ABSTRACT: Ground deformation induced by tunnelling in shallow sandy ground can be reduced by placing
some reinforcements such as facebolts and forepoling bolts from the tunnel. A series of centrifuge tests have been
carried out in order to investigate the ground deformation pattern during tunnel excavation with reinforcements.
Three dimensional numerical analysis of the problem was also performed using FLAC3D and the simulation
results show good agreement with the centrifuge data.
1
INTRODUCTION
Tunnel reinforcement has been applied to bored tunnel

excavation in order to keep the cutting face stable, to
reduce ground (sub)surface settlements and to avoid
any adverse influence on adjacent structures. Forepoling and facebolts are the two most popular tunnel
reinforcements; the former is often used in European
countries, whereas the latter is frequently applied in
Asian countries. However, the specifications of using
them are based on local and empirical designs or on
experience of past constructions in similar ground
conditions.
In order to excavate a larger tunnel under poor
ground conditions safely, it is necessary to establish
a new design method for tunnel reinforcements such
as forepoling and facebolts. In particular, in order to
evaluate the relative merit of these two techniques for
ground deformation control, it is important to compare
them using the same modeling techniques (centrifuge
tests or numerical analysis) under the same ground
conditions.
2
extend to the ground surface. The centrifuge tests by

Leca and Dormieux (1990) show that the failure lines
reached the ground surface when C/D = 1. These two
studies indicate that ground deformation pattern, when
C/D is less than 1.0, is different from that C/D exceeds
1.0. Hence, it was decided to perform centrifuge
experiments with a tunnel model of C/D = 1.0.
2.1
Models
A schematic view of the model is shown in Figure 1.

The strong box shows the half of the prototype so that
the ground deformation changes could be observed
through a perspex window that is installed on the
longitudinal section of the model. The ground deformation was analyzed using PIV program developed by
White & Take (1996).
CENTRIFUGE TESTS
In this study, the effect of tunnel reinforcements on

ground deformation in shallow tunnels was investigated. Chambon and Cort (1994) performed centrifuge modeling of tunneling in sandy ground and
showed the minimum pressure to support a cutting
face was independent of cover diameter ratio (C/D, C:
cover, D: tunnel diameter). They also showed, when a
model tunnel was installed with C/D = 4 and P = 0.1D
(P: unsupported length), the failure lines extended to
the height of about 2.5D from the crown and did not
Figure 1. Schematic view of strong box and model tunnel.
635
Table 1.
Matrix of reinforcement bolts.
Test
Type
Arrangement*
Number
KDC10
KDC04
KDC05
KDC06
KDC07
No reinforcement
Facebolts
Facebolts
Forepoling
Forepoling
FB01
FB02
FP01
FP02
0
14
14
14
28
* Each pattern is displayed in Figure 2.
The model box was filled with dry Leighton Buzzard Fraction E Silica sand with the relative density
of 87% (2%), corresponding to a unit weight of
15.9 kN/m3 .
The model tunnel, of diameter D = 100 mm (7.5 m
in prototype), is semicircular and the depth to the tunnel crown C was equal to the diameter D (C/D = 1.0).
The excavation of the tunnel was simulated by decreasing the internal pressure of a rubber bag placed at the
tunnel face. The bag was covered with an alminium
rigid lining, which was installed at a distance of P
(=0.2D) behind the tunnel heading. The internal pressure was reduced from 100 kPa to tunnel collapse
pressure. The centrifuge tests were performed at 75 g.
A series of five tests were carried out as listed in
Table 1. The model reinforcement bolts, which were
made of alminium, were installed perpendicularly to
the tunnelling head during the sand-pouring. They
were coated with the same sand as used in the tests,
the outer diameter of them were 2.4 mm (180 mm in
prototype). The forepoling model bolts were attached
to the model liner with glue. The number, the length
and the arrangement of the reinforcements were varied
as shown in Table 1 and Figure 2.
2.2 Tunnel collapse
The model tunnel without tunnel reinforcement
(KDC10) collapsed at the support pressure of 3.1 kPa,
which agrees with the past centrifuge results by
Chambon & Cort (1994). On the other hand, for the
tunnels with reinforcements (KDC04-07), tunnels collapsed at lower pressures (2.22.7 kPa). This shows
that not only facebolts but also forepoling contributes
to reducing the minimum support pressure, but both
techniques did not dramatically decrease the pressure
required to keep the face stability.
The tunnel collapse mechanism on the longitudinal section is shown in Figure 3. When there is no
reinforcement (KDC10), the front slippage line started
at the bottom of the tunnel, extended upwards with
a quasi-logarithmic curve, and then reached to the
ground surface vertically. The line behind the tunnel was nearly vertical but a little inclined backwards.
The failure mechanisms using facebolts are shown by
KDC04 and 05 in Figure 3. The distinct difference
from the non-reinforcement case was found at the front

failure line, which started from some point at the upper
face and then did not extend ahead of the face but
extended upwards almost vertically. This showed that
facebolts were effective to improve the face stability.
In other words, this change in failure mechanism led
to the reduction of the collapse volume. When facebolts were installed only at the upper face (KDC05),
a chimney-like collapse was observed. This may be
because the tunnel collapse was more dominated by P,
that is, the stress relaxation at the crown rather than
the height of the model tunnel, H.
When forepoling was introduced in a sparse manner
(KDC06), the front slippage line was similar to that of
non-reinforcement case (KDC10), but the back slippage line did not develop outward but inward toward
the tunneling direction. This back line pattern was similar to that when a denser pattern of forepoling was
introduced in KDC07. However, the geometry of the
front line was totally different. In KDC07, the slippage line developed from the middle of the heading,
and extended up to the horizontal line where forepoling
bolts are embedded. Then the line suddenly developed
vertically upwards to the ground surface.
When forepoling bolts are densely installed, they
divide the surrounding ground into two zones; (a) the
outside zone of invisible arch consisting of forepoling
bolts, and (b) the inside zone of the forepoles. The
mechanisms of collapse of the two zones have to be
considered separately. It appears that the collapsed area
shifted forward in the longitudinal direction.
2.3
Changes in displacement vectors with the

decrease in tunnel support pressure
The ground deformation in sandy ground is quite small

even when tunnel support pressure decreases to about
half of the initial pressure. However, once it starts to
develop, it abruptly increases and then reaches the
collapse rather instantaneously. Therefore, in the past,
it has been difficult to obtain the displacement vectors in sandy ground as the tunnel support pressure
decreases. The PIV analysis, developed by White et al.
(2003), was used to monitor subtle changes in ground
deformation in sandy ground.
Figure 4 shows that the changes in the distribution
of displacement vectors on the longitudinal section
in KDC10 as the tunnel support pressure decreases.
The displacement vectors were difficult to be detected
even by using the PIV analysis until the internal stress
was unloaded to around 25 kPa. Initially, the region
deformed by the stress release was widespread. As
shown in Figure 4, with decrease in the tunnel support pressure, the deformation became more localized
around the face rubber bag. The first large ground
movement was observed at 5.4 kPa and it was an earlike shape on the longitudinal section. The observed
636
Figure 2. The arrangement patterns of bolts.
shape is similar to the three dimensional tunnel failure

mechanism proposed by Leca & Dormieux (1990).
2.4
Displacement vectors near collapse in

centrifuge tests
Figure 5 shows the displacement vectors at pressure

close to the tunnel collapse, which varies from 3.5 to
5.0 kPa.
The distributions are comparable to the failure
shapes shown in Figure 3. It was obvious from KDC04
and KDC05 that facebolts were effective to reduce the
front area affected by tunnel excavation and also the
ground movements, in particular, for the horizontal
one. In the forepoling cases (KDC06 and KDC07), it
appears that the ground moved under different mechanisms at the inside and outside of the half arches
created by forepoling bolts as shown in Figure 1 and
Figure 2. Some continuity between the inside and the
outside is observed in KDC06, but they are discontinuous in KDC07. Hence, it was concluded that forepoling
bolts can be effective to reduce the displacement outside the forepoling arch as long as they are installed
densely enough to divide the surrounding ground into
two zones.
Figure 6 shows the distribution of horizontal displacement along the cutting head at different internal
pressures for KDC10, KDC04 and KDC07. Results
show that facebolts (KDC04) contributed to the reduction of face extrusion. For the case of densely installed
forepoling bolts (KDC07), the effect to reduce the
face extrusion was not so apparent as that with facebolts, but the bulging pattern changed. That is, the
maximum extrusion was found at the location beneath
the crown, although it was found at about 1/4 the
height of the face from the crown in KDC10 and
KDC04. Ground deformation started at 2025 kPa for
all cases, but the deviation from the non-reinforcement
case (KDC10) to the reinforcement cases (KDC04 and
KDC07) became evident when the face pressure was
1015 kPa.
637
In actual construction, great care must be taken in

order to avoid fatal ground (sub)surface settlement
above a tunnel, and hence it is essential to study how
the settlement trough develops with the decrease in the
tunnel pressure especially at locations just above the
tunnel crown.
Figure 7 shows the subsurface settlement at just
above the tunnel crown in KDC10, KDC04 and
KDC07. These troughs appeared when the face pressure was 2025 kPa for all cases, but the differences
among them became apparent when the face pressure
was 1015 kPa. At around 5.56.5 kPa, the maximum
settlements in KDC04 and KDC07 were about half
Figure 5. Displacement vector distribution on the longitudinal section.
Figure 3. Tunnel failure patterns on the longitudinal section.
Figure 6. Distribution of face extrusion at the face.

Figure 4. Displacement vector changes with the decrease in
tunnel support pressure.
as large as those in KDC10. Hence, both facebolts
638
Figure 8. Geometry of 3D numerical analysis model.

Table 2.
Model
Soil properties for test simulation with FLAC3D.

Youngs modulus E
c(Pa) ( ) ( )
Mohr-Coulomb-1 E = p * 1400
0.1 40 0, 15
p = (v + h + h )/3
Mohr-Coulomb-2 h = v /(1 ) 0.1 See Figure 9
* Mohr-Coulomb model without strain softening/hardening
model
** Mohr-Coulomb model with strain softening/hardening
model
tests; swing-up to 75 g, and then decrease of the internal pressure inside the rubber bag. The tunnel lining
was assumed to be rigid.
Figure 7. Subsurface settlement just above crown.
and forepoling were effective to reduce ground settlements. The shapes of the troughs are sharper in
KDC10 and KDC04 than in KDC07. The maximum
settlement in KDC4 was positioned behind that in
KDC10. The trough in KDC07 was wider than the
troughs observed in KDC04 and 10, and, as a result,
the position of maximum settlement shifted ahead of
the maximum settlement occurred in KDC10. Densely
installed forepoling bolts were capable of reducing
the influences from the stress release at both the face
and the crown, while facebolts only contributed to
counteract the effect of the stress release at the face.
3
3.1
3.2 Soil property

Mohr-Coulomb without strain softening/hardening
model (MC), one of the standard and basic models,
was used. Two dilation angles were used as shown in
Table 2.
In addition, in order to simulate more accurately
the results from the centrifuge tests, Mohr-Coulomb
with strain softening/hardening model (SSH) was also
used. The changes in parameters with plastic shear
strain were derived from the triaxial tests by Coelho
et al.(2007). As shown in Figure 9, the simulation
of triaxial test results using FLAC3D showed good
agreement with the test performed under the initial
confining stress of 120 kPa.
SIMULATION MODELING AND ANALYSIS

3.3 Simulation results for the non-reinforcement
case
Modeling
In order to simulate the centrifuge test results, 3D analyses were performed using FLAC3D. The geometry
was identical to the internal size of the strong box used
in the centrifuge modeling tests; 400 200 300 h in
mm, as shown in Figure 8. For the boundary conditions at side walls, roller conditions were applied. The
analysis basically followed the order of the centrifuge
Figure 10 shows the development of the maximum

face extrusions with the decrease in the tunnel support
pressure. The centrifuge exhibited the entire collapse
at 3.1 kPa. The prediction with the SSH model was in
good agreement with the experimental results. When
the MC models were used, the deviation from the
experimental results starts at around 3040 kPa and
639
Figure 9. Simulation results of triaxial test at 3 = 120 kPa.
Figure 11. Simulation of the face extrusion bulges in

KDC10 when the support pressure was 8.09.0 kPa.
Table 3.
Input parameters for facebolts.
Model
Density
Youngs modulus
Poissons ratio
Pile
2700(kg/m3 )
7*104 (MPa)
0.2
Figure 10. Simulation of the maximum face extrusion in

KDC10 with 3D numerical analysis with the Mohr-Coulomb
models with/without strain softening/hardening model.
then abrupt increase in face extrusion was found at

around 15 kPa.
This is because the SSH model is able to simulate the
plastic behaviour at and after small strain levels, corresponding to the face extrusion larger than 1.5 mm. As a
result, the SSH model can follow the gradual increase
in face extrusion with the decrease in tunnel support
pressure.
Figure 11 shows the simulation results of the distribution of face extrusion in KDC10 when the tunnel
support pressure was 8.09.0 kPa. Both constitutive
models were successful in estimating the extrusion
curve just before tunnel collapse, but, in order to simulate the deformation pattern observed in the experiments more precisely, the SSH model was found to be
more appropriate than Mohr-Coulomb model.
However, there was a difference in position where
the maximum extrusion occurred in Figure 11. In the
numerical analyses, the peak was found at the middle and this location is lower than the location of
the peak observed in the centrifuge test. This may
be due to the difference in the shape of the excavation ranges in between centrifuge tests and numerical
analyses. The model ground in centrifuge tests can
move smoothly at the corner near the crown and the
face because the rubber bag is flexible enough to be
smoothly deformed. However, as for the numerical
Figure 12. Changes in shear stiffness and yielding strength

of soil-bolt interface with overburden pressure.
analyses, rectangular meshes might prevent the model

ground from extruding inwards smoothly.
Further improvement in the numerical analysis (e.g.
mesh making) is required in order to illustrate not only
the development of the maximum extrusion but also
its distribution along the face.
3.4 Reinforcing effects of facebolts on reducing
face extrusion
The reduction of face extrusion by facebolts was also
simulated. The SSH model was adopted for both simulations. The interface property between ground and
bolt surface, which is shown in Table 3 and Figure 12,
was derived from the pull-out tests reported in Date
et al. (2007). Figure 13. shows the simulation results
of KDC04. Results from KDC10 are also presented
for comparison purpose.
The deformation pattern simulated by the numerical
analysis was similar to that of the centrifuge test, but
the magnitude predicted by the numerical analysis was
640
Figure 13. Simulation of face extrusion in KDC4 with 3D

numerical analysis with the SSH model.
smaller than the centrifuge data. This may be due to

the mesh problem mentioned above, the difference in
the soil-bolt interaction properties and the difference in
soil properties during between loading and excavating.
Further investigation is needed.
4
CONCLUSIONS
A series of the centrifuge tests showed that introduction of facebolts and forepoling bolts for tunneling in
shallow sandy ground yielded different shapes of tunnel collapse and contributed to some reduction in the
tunnel support pressure to keep the cutting head stable.
These techniques were also effective in reducing the
vertical settlement at locations just above the tunnel
crown. Facebolts were able to reduce the risk of face

extrusion and hence can make the ground ahead of
the face stiffer. Forepoling bolts can divide the ground
around the tunnel face into two zones; the inside and
the outside of the arch of forepoling bolts.
Numerical analysis results show that the SSH
model (Mohr-Coulomb model with strain softening/hardening) gave better match to the centrifuge
data than the MC model (Mohr-Coulomb model without strain softening/hardening). However, in order to
simulate more realistic behaviour observed in the centrifuge tests, further investigation of the settings of
numerical analysis such as mesh-making and soil-bolt
interaction properties needs to be conducted.
REFERENCES
Chambon, P. & Cort, J.F. 1994. Shallow tunnels in cohesionless soil: stability of tunnel face. Journal of Geotechnical and Geoenvironmental Engineering, ASCE, 120(7):
11501163 1994.
Coelho, P.A.L.F. 2007. In situ densification as a liquefaction
resistance measure for bridge foundations. PhD Thesis,
Cambridge University, UK., 2007.
Date, K., Mair, R.J., Soga, K. Centrifuge tests for tunnel reinforcement. Soils and Foundations, 2008. (in preparation)
Leca, E., Dormieux, L. 1990. Upper and lower bound solutions for the face stability of shallow circular tunnels in
frictional material. Gotechnique 40(4), 581606 1990.
White, D., Take, A. Bolton, M.D. 2003. Soil deformation
measurement using particle image velocimetry (PIV) and
photogrammetry. Gotechnique 53(7), 619631 2003.
641
Mechanical behavior of closely spaced tunnels laboratory

model tests and FEM analyses
J.H. Du & H.W. Huang
ABSTRACT: Reduced-scale model tests and corresponding finite-element analyses were performed to investigate the effect of different excavation sequences for twin tunnels under the unsymmetrical pressure in weak
rock. In these tests, the alternative sequences of the excavation of the left tunnel first and then the right one, or
the right first and then the left, were simulated. The displacement of ground and the earth pressure in the rock
were measured. Particular attention was paid to the different behaviors of the displacements and the stresses
caused by the different excavation sequences. The result of the present study indicates that the state of stresses,
the displacements around the tunnels and the ground surface settlements are all different when the excavation
sequence is alternated. When the right one is excavated firstly, the stratum condition is more deteriorated. It can
be concluded that the sequence of the left first and then the right one is better for such twin tunnels.
INTRODUCTION
With the development and the upgrade of infrastructures such as highway, subway, railway, and many other
facilities, tunnel constructions are gradually increasing in the recent years. The preference in China, and
as well as in other countries, is to use twin tunnels for
the new transportation lines rather than a single larger
diameter tunnels, when the space is limited.
Model test and numerical simulation are two key
methods to investigate the tunneling problems. In the
past two decades, the rapid advances have been made in
tunnel model tests, many of which are investigations on
the twin tunnels. Dhar et al. (1981) performed the fracture pattern around twin openings in weak materials
of sand wax and sand wax mixtures with the plaster of Paris under controlled loading conditions. Twin
openings for different orientations in respect to loading directions were studied. Adachi et al. (1993) also
used model tests to analysis the interaction between
twin tunnels. The tunnel excavation was simulated
by tailor-made diameter reducible device. Chu et al.
(2006) performed model tests of twin circular tunnels
in homogenous material, two-layered formations, and
three-layered formations to understand the mechanical behavior of a twin-tunnel of circular cross section
in multi-layered formations, and a two-dimensional
numerical simulation was also developed.
Using 2D and 3D numerical simulations, Ghaboussi
(1977), Soliman (1993), Addenbrooke (1997), Ng
Figure 1. Typical cross section of Pingnian Tunnel.
(2004) et al. investigated some aspects of interaction

mechanisms.
Pingnian Tunnel is a typical twin-tunnel, which was
constructed in accordance with the principles of the
New Austrian Tunneling Method in weak rock, being
part of Luofu expressway in Yunnan Province. Tunnel
width is 12.5 m, and the length is about 360 m. Spacing
between two tunnels changes from 18 m (at Luochunkou site) to 25 m (at Funing site). The tunnels are under
the inclined ground surface and the two tunnels are not
constructed in the same level, and the vertical distance
between them is from 3 m to 4 m. The typical cross
section is shown in figure 1.
643
Table 1.
tests.
Properties of the ground material for the model
Unit weight (KN/m3 )

Youngs Modulus E (MPa)
Cohesion c (KPa)
Friction angle (deg)
Figure 2. Sketch of steel frame with model tunnel

(Unit: mm).
So far, the behavior of surrounding ground during the single tunnel excavation has been extensively
investigated, however, the case that a tunnel is driven
paralleling to another adjacent tunnel is of interest
to tunnel engineers. During the Pingnian Tunnel construction, a collapse happened at the left line portal,
and the slip also affected the adjacent portal of right
line because of the small space between the two tunnels. In order to study the actual interaction between
two tunnels, two-dimensional laboratory model tests
were performed to capture the earth pressure transfer
and the displacement of surrounding rock during construction. The particular attention was paid to the evolution of the displacements and stresses caused by the
different excavation sequence. The numerical simulations by the FEM were also conducted for the problem
to compare the analytical results by model tests.
2
SIMULATION METHODS
2.1 Two-dimensional plane strain test

Although actual tunneling is a three dimensional problem and actual ground is heterogeneous as well as
anisotropy in nature, in this study, plane strain tests
were conducted in a homogeneous ground by simplification. The box used in tests was made of a
steel frame, as shown in figure 2, with dimensions of
1.6 m 1.2 m 0.4 m. The four sidewalls of the test
26.9
38
25
39
box were assembled using steel sheets. For the case of

the observation of the ground movement patterns during testing, two transparent Perspex plates were used
at the front and back sides.
The tests were conducted according to the typical
cross section in the steel frame as shown in figure 2.
The tunnel width is 156 mm, the horizontal distance
between two tunnels is 225 mm, the vertical distance
is 500 mm under inclined ground with angle 18 , and
the cover depths of left and right tunnel are 212 mm and
138 mm respectively. All dimensions are controlled by
the geometrical similarity ratio of 1/80.
The ground material used in the model test was composed of barite powder, sand and plaster mixed with
water, and the quality proportion was achieved through
a series of material tests. The main parameters of the
ground material are shown in table 1, determined by
the following expressions according to the similarity
theory.
where C , C , C , CE , Cc , Cl are the similarity ratios

for friction angle , stress , youngs modulus E,
cohesion c, unit weight and geometrical dimension l, which represent relationship for the parameters
between the model test and the real situation.
The excavation was simulated by removing the
ground inside the excavation zones manually, with the
two different excavation sequences conducted: the left
tunnel excavated first and then the right one (named
test A), and the right first and then the left (named
test B). A monitoring program was utilized to record
the process of the ground movements and the earth
pressure changes around the tunnels. Monitoring items
included earth pressure cells, deep rods and micrometer gauges, and the locations of items shown in figure 3
and figure 4. All monitoring items were attached to the
high-speed instrument of static strain gauge (YE2539),
collecting data at any interval operator set.
644
Figure 5. Surface settlement curves (obtained in model

test).
Figure 3. Locations for displacement gauges.
Figure 4. Locations for earth pressure gauges.
2.2
Numerical simulation
Corresponding numerical simulation was performed

to compare the above experimental tests. The simulations were based on the model test conditions
and the material properties. The finite element software Msc.Marc was adopted for the simulations. In
the analyses, the ground material was assumed to be
elastic-plastic confirming to the Druker-Prager failure criterion together with the associated flow rule.
The analysis consisted of two phases: the first phrase
is to create an initial geostatic condition and the second
phase to simulate the excavation process. During the
first phase, the initial geostatic condition was achieved
by applying gravity forces to all ground elements. In
the second phase, the excavation process was simulated by removing the elements inside the excavation
zone. The simulation of the test A and the test B
performed in laboratory model tests were conducted
accordingly.
3
3.1
RESULTS
Displacement
Figure 5 is a surface displacement-time diagram,

which is obtained in the model test. From the figure we
can find that the settlement of every test point increases

step by step along with the advancement of excavation. A substantial increase of the displacement occurs
when excavation section passes the test section. The
settlement develops continuously after excavation finished and reaches the steady-state in a period of time.
The creep displacement is the major part during the
period of the excavation termination, accounting for
about 10%30% of the total displacement.
There is still some ground surface settlement (at
G2, G4) increasing during the excavation process of
adjacent tunnel, which reveals the existence of an interaction between the twin tunnels. G1 located above and
left to G2, whose horizontal distance to the left wall of
the left tunnel is one time of width of the tunnel, doesnt
undergo a distinct displacement during the excavation
of the right tunnel while it does during the excavation of the left one. The displacement of the lower G5,
located in the right to the G4, is totally different, which
is mainly affected by the excavation of right tunnel. For
the spot G3, located at the surface above the pillar centerline, the settlement is still in the development state
and the excavation of both the right and the left tunnels
will produce effect on it; the two abrupt changes of the
displacement happened in the excavation section passing the test one. The displacement change of the spot
G3 appears smaller than the displacement of the spot
G2 and G4.
We can make a conclusion from the above: the influence of tunnel excavation decreases with the increase
of the distance to the excavated tunnel. The zone influenced mostly is those above the tunnels. Under such
leaning stratum, the influence of the excavation to
rocks in one time of the width of the tunnel region
still exists, but it is small when the distance between
the tunnel and the rock is more than three times of
the width of the tunnel. Excavations of left and right
tunnel all have effect on surface above the pillar.
The comparison between deep settlement (settlement of the inner stratum, indicated by D1, D2 and D3,
645
Figure 8. Transverse settlement trough corresponding to

different excavation sequences (obtained in FEM analyses).
Figure 6. Surface and deep ground displacement curves
(obtained in model test).
Figure 7. Surface and deep ground displacement curves

(obtained in FEM analyses).
located above the left tunnel crown, in the pillar center

line parallel to the right tunnel spring line and above
the right tunnel crown respectively) and corresponding
surface settlement is shown in figure 6. The development trends are coincident, however, deep settlement
occurs prior to that of the surface, and its displacement rate is also relatively bigger. Time to steady state
needed by the deep displacement is shorter than that
of the surface settlement. When it comes to compare
two tunnels, the displacement rate of the left tunnel is
bigger than the right one.
In the test A, settlements of all measure points are
more than those got in the test B. The maximal displacement is 35.9 mm when the left tunnel excavated
first and then the right one, and it is 44.4 mm when
the right one first and then the left. So it indicates
that the left tunnel excavated first can control surface displacement more effectively. The comparison
of surface settlements between two opposite construction sequences achieved in numerical simulations are
shown in figure 7. It reaches the same conclusion

that the ultimate surface settlement occurred by the
right tunnel excavated first is bigger than that occurred
in the reverse excavation sequence. The evolution law
of the displacement rate is the same of the settlement.
So it is not beneficial for the tunnel stabilization to
adopt the construction sequence with the right tunnel
excavated before the left tunnel under such locations
as the Pingnian twin tunnel under the inclined stratum.
The whole surface settlement curves can be
obtained through FEM analyses, shown in figure 8.
Based on these curves it can be concluded that, with the
right tunnel finished, the excavation of the left tunnel
(the curve B in figure 8) will generate more displacement than that caused by the excavation in green field
(the curve A in figure 8). The disturbance due to the
right tunnel construction lowers the stiffness of the
rock around the left tunnel and when the left is constructed, the rock in the previously disturbed zone is
disturbed again, and it will move more than what is
expected. So, the existing of the right tunnel is undesirable for the left one. However, when there is the
adjacent tunnel existing, additional ground surface settlement caused by right tunnel excavated (the curve D
in figure 8) is less than that caused in green field (the
curve C in figure 8). The existing of the left tunnel
which is higher than the right one is an advantage for
the right tunnel to a certain extent.
3.2 Earth pressure

Earth pressure near the crown and the invert (R1 is
above the crown and R2 is under the invert, refer to
figure 4) of the right tunnel are given in figure 9. There
is the same development rule for the earth pressure
of R1 and R2; the pressure increase a bit due to the
previous tunnel excavation, and decrease rapidly as
excavation section passes the test section because of
the load release caused by removing rock of the tunnel
space.
646
Figure 9. Earth pressure near crown and invert of right

tunnel (obtained in model test).
Figure 10. Earth pressure at crown and invert of the right

tunnel.
The pressure change at R2 near the crown is bigger than that at R1 near the invert. The phenomena
should be associated with the locations of two measuring points, the invert is deeper than the crown and
initial stress near the invert is bigger than that near the
crown. The same conclusion can be drawn by numerical simulation and the results are shown in figure 10.
The stress changes caused by the excavation of the
right tunnel in green field are more than that caused
with the existing left tunnel. And it comes to opposite
conclusion for the left tunnel. So, depending on the
stress changes, the same conclusion can be drawn: the
existing of the right tunnel is undesirable for the left
one; and the existing left tunnel is beneficial to the
right one.
CONCLUSION
The reduced-scale laboratory tests and their corresponding numerical analyses were performed to
investigate the transfer of the earth pressure and the

displacement of the surrounding rock during the construction. At the same time the effect of different
excavation sequences was analyzed.
The degree of the interaction between the two tunnels is proportional to the distance between them. The
changes of stresses and displacements of the rocks
along the tunnel are mainly affected by its excavation while the effect caused by the adjacent tunnel is
small. The interaction effects on the two tunnels are
different: the existing of the right tunnel deteriorates
the properties of rocks around the left tunnel, which
is harmful to the left tunnel; whereas the existing left
tunnel reduces the asymmetric degree, and therefore
benefits to the right tunnel.
Stress decreases rapidly with the excavation of the
tunnel, which causes settlements of the surrounding
rocks and consequently expands to surface.
Displacements of the rock increase gradually along
with the excavation advancement, and the main portion occurs when the excavation section passes the test
section.After the excavation finished, a creep displacement takes place, which accounts for about 10%30%
of the total displacement.
The difference of the excavation sequence causes
the different development rule of the stress field and
the displacement field. The values of displacements
and stresses increase on adopting right tunnel excavated firstly. For the tunnels like the Pingnian tunnel,
which is a shallow tunnel under unsymmetrical pressure with the vertical distance between the tunnels, it
is profitable to excavate the left tunnel firstly.
REFERENCES
Adachi, T., Kimura, M. & Osada, H. 1993. Interaction between multi-tunnels under construction. Eleventh
SoutheastAsian Geotechnical Conference, Singapore, 48
May 1993: 5160.
Addenbrooke, T.I. 1996. Numerical analysis of tunnelling
in stiff clay. Ph. Degree thesis, London: University of
London.
Chu, B.L., Hsu S.C., Chang Y.L. et al. 2006. Mechanical
behavior of a twin-tunnel in multi-layered formations.
Tunnelling and Underground Space Technology.
Dhar, B.B., Ratan, S., Sharma, D.K. et al. 1981. Model study
of fracture around underground excavations. Proceedings
of the International Symposium on Weak Roc: 267271.
Ghaboussi, J. & Ranken, R. E. 1977. Interaction between two
parallel tunnels. International Journal for Numerical and
Analytical Methods in Geomechanics. 1:75103.
Ng, C.W.W., Lee, K.M. & Tang, D.K.W. 2004. Threedimensional numerical investigations of new Austrian
tunnelling method(NATM)twin tunnel interactions.
Geotechnique. 41:523539.
Soliman, E., Duddeck, H. & Ahrens, H. 1993. Two and
three dimensional analysis of closely spaced double-tube
tunnels. Tunnelling and Underground Space Technology.
8(1):1318.
647
Stability analysis of masonry of an old tunnel by numerical modelling and

experimental design
J. Idris & T. Verdel
Laego Ecole des Mines de Nancy, Parc de Saurupt, Nancy Cedex, France
M. Alhieb
Ineris Ecole des Mines de Nancy, Parc de Saurupt, Nancy Cedex, France
ABSTRACT: The present paper proposes two numerical models of an old tunnel supported by masonry; these
models were developed by the well-known Universal Distinct Element Code (UDEC). A masonry mechanical
behaviour analysis and numerical simulation of the masonry ageing phenomenon will also be addressed by
means of an experimental design to study the influence of masonry block physical properties on the mechanical
behaviour of tunnel support structure.
INTRODUCTION
strategies for masonry structures can at present be

identified (see figure 1):
Masonry vaults are one of the most familiar structural

shapes present in architectural heritage worldwide.
Historical domed buildings arched stone bridges and
vaulted tunnels are among the most famous examples.
Over the last few years, the development of numerical tools in the field of structural analysis has enabled
researchers to establish approaches for numerically
modelling masonry structures, yet an analysis of
the mechanical behaviour of such blocks and joints
structures remains challenging due to the extent of
correlated factors. This paper seeks to study tunnels
masonry behaviour used in old tunnels by numerical
modelling and experience design.
2
1 Detailed micro-modelling: blocks and mortar in the

joints are represented as a continuum, whereas the
unit/mortar interfaces are modelled by discontinuous elements;
2 Simplified micro-modelling: geometricallyexpanded continuum units, with discontinuum
elements incorporating the behaviour of both mortar joints and interfaces;
3 Macro-modelling: all three principal features of
structural masonry are represented by an equivalent
continuum.
MASONRY STRUCTURES NUMERICAL

MODELLING APPROACHES
Several modelling approaches to masonry structures

(continuous and discontinuous modelling) are currently under development by several research teams. A
number of bridge arch models have been proposed to
study the behaviour and certain instability problems,
such as collapse (Ford et al., 2003; Sumon et al., 1995;
Hughes et al., 1998; Bicanic et al., 1995; Brookes
et al., 2004). Many historical masonry construction
simulations and numerical analyses were presented
by (Valluzzi et al., 2004; Loureno, 2001; Giuriani
et al., 2001; Bicanic et al., 2002). Yet no single publication actually focuses on or addresses the analysis of
old tunnel masonry structures. Three basic modelling
Figure 1. Basic approaches to masonry structures.
649
Figure 3. Sample disorder types within old tunnels (CETU

Tunnel Study Centre, 2004).
Figure 2. French tunnel classification (Idris et al., 2007).
3 TUNNELS AGEING PHENOMENA

Old underground structures, especially tunnels, display specific characteristics regarding behavioural
evolution over time. The infrastructure environment,
surrounding soils and construction materials used all
contribute to this evolution. Consequently, several
types of disorders may appear and develop in these
underground structures due to specific ageing processes that, in reality, are complex phenomena. One
impact is the alteration in mechanical properties of
the various construction materials.
A majority of the worlds tunnels are currently more
than 100 years old; these would all be considered as
ancient infrastructure. Figure 2 illustrates the classification of tunnels in France, where the oldest one
exceeds 180 years.
The majority of old tunnels are supported by
masonry elements. The type of masonry support or
lining depends upon utilisation of the high compressive strength in the stones, which explains the vaulted
section shape of old tunnels supported by masonry.
Apart from the environment and evolution in surrounding soil and in the absence of an effective
isolation system for such underground structures, subsoil water can easily penetrate the masonry joints and
circulate within.
Over time and in the presence of other aggressive
ambient factors, several physical, chemical and biological processes may develop inside the masonry structure; this phenomenon and its impacts are collectively
called the tunnel-ageing phenomenon.
As a result, various types of disorders appear
inside old tunnels; these would include: longitudinal
or transverse structural cracks, convergence and partial masonry collapse. Figure 3 provides some images
of such disorders.
NUMERICAL MODELLING OF ANCIENT

TUNNELS
A tunnel masonry structure is a discontinuous medium

consisting of blocks bonded to each other by mortar;
Figure 4. A typical old tunnel cross-section (extracted

from Kerisel, 1975).
in addition, such a structure forms an interface with

the surrounding soil. The Distinct Element Method
(DEM) is a suitable technique for modelling these
structures. By means of the Universal Distinct Element Code (UDEC), two simplified micro-models of
an ancient tunnel have been derived. The geometric
and geomechanical properties of both models were
inspired from previous cases of ancient tunnels available in the old tunnel sub-database (Idris et al.2004,
2007) as well as from other bibliographical sources,
including Kerisel, 1975 (see Figure 7, in which the
thickness of the masonry support structure equals
80 cm).
The two representative models are positioned at a
shallow depth of 20 m; they both display a vaulted section shape. In the first model, the masonry-supporting
section consists of a regular rectangular and square
limestone blocks (Fig. 5a). In the second, the masonry
consists of two layers of limestone blocks: regular and
irregular (Fig. 5b). Masonry blocks are bonded by a
lime mortar. The masonry support thickness is 80 cm
and the sidewall height in each model amounts to 3 m.
All other geometric details are shown on Figure 5.
By taking into account model section symmetry, only half of each set-up needed to be modelled. Figure 6 shows the tunnel position within the
650
Figure 5. Two proposed models (a and b) for old tunnels.
surrounding soil, along with the selected boundary

conditions, the chosen dummy joints and the mesh.
Dummy joints were introduced in order to obtain a
suitable grid around the excavated tunnel.
The soil surrounding the tunnel consists of a homogeneous mix of clay and sand. Table 1 lists the basic
mechanical properties assigned to the surrounding
soil, masonry and masonry joints, based on the work by
(Verdel et al., 1999), (Hoek, 2000) and (Janssen, 1997).
Calculations were carried out in plane strain: the
soil and masonry follow a perfect elasto-plastic MohrCoulomb plasticity criterion. The calculation step
proceeded by two main stages: model consolidation in
the initial stress condition prior to tunnel excavation;
and tunnel excavation and simultaneous installation
of masonry support. The calculation could then be
continued until reaching model equilibrium.
5
MECHANICAL BEHAVIOUR ANALYSIS

OF MASONRY BLOCKS
Figure 8 presents Mohr circles for every grid zone of

the masonry block used in the initial model (according
to which, every masonry block is divided into 2 or
Figure 6. Tunnel position, boundary conditions and mesh.
4 grid zones). In the initial model, no block stress

exceeds the Mohr-Coulomb failure criterion (as shown
on Figure 7), and all masonry blocks display perfectly
elastic behaviour. Based on the stress distribution of
masonry blocks, we are able to approximate the critical values for masonry block cohesion, tensile strength
and friction angle, all of which can modify the masonry
behaviour from elastic to plastic.
This analytical approach has enabled us to define
the variation range for these three parameters over
651
Table 1. Selected mechanical properties of the surrounding soil, masonry and masonry joints (M: Volumic mass; E:
Youngs modulus; : Poissons ratio; C: Cohesion; : Friction angle; Tr: Tensile strength Jkn, Jks: Normal, Tangential
joint stiffness; JC: Joint cohesion; J: Joint friction angle;
JTr: Joint tensile strength).
Surrounding soil
Parameter Unit
M
E
Tr
Kg/m3
MPa
MPa
MPa
Masonry
Value
Parameter Unit
1900
200
0.3
0.50
20
0.10
M
E
Tr
Value
Parameter Unit
Value
Kg/m3 2100
MPa
6000
0.2
MPa
3
30
MPa
1
Figure 8. A sample of numerical simulation results for the

first model (test, cohesion, tension and friction).
Masonry joints
Parameter Unit
Jkn
Jks
JC
GPa/m
GPa/m
MPa
150
69.7
1.2
J
JTr
MPa
Value
25
0.4
Figure 7. Mohr circles for masonry block grid zone stresses

in the initial model.
which it is appropriate to study how the three chosen parameters influence the overall behaviour of
masonry structures. As an example, the decrease in
both cohesion and friction angle can serve to significantly increase the number of plastic zones within the
masonry structure.
PROPOSED EXPERIMENTAL DESIGN

FOR MASONRY SUPPORT AGEING
SIMULATION
chemical and biological combined processes. A number of physical phenomena, such as dilation, contraction, freezing, hydration and desiccation cycles,
potentially lead to a considerable evolution in material
mechanical properties. Construction materials ageing
processes detail were explained by (Idris et al. 2007).
The behaviour of masonry is dictated by the physical
and mechanical properties of both blocks and joints.
Our survey focused on evolution in the mechanical
behaviour of masonry blocks due to ageing; special
attention was therefore devoted to the parameters that
define block failure, such as cohesion, tensile strength
and friction angle.
To evaluate the influence of each chosen factor on
masonry structure behaviour, it proved necessary to
observe significant changes in model behaviour once
factor values had been changed. The response factor
selected herein is the number of blocks that change
status from elastic to plastic behaviour.
For this purpose, a complete factorial design was
proposed; this three-level design is written as Kn factorial design (with K = 3: the studied factor number,
n: level number). This nomenclature means that 3
factors are considered, each one at 3 distinct levels
(Barrentine, 1999).
Consequently, a complete factorial design with 27
experiments was proposed. Table 2 contains all of the
experimental results (i.e. changed experimental factors and observed responses), and Figure 8 provides
some of the results for both models.
We have to indicate that the obtained results from
the two developed models are quite similar.
Construction materials are submitted to various modifications and degradations under several physical,
RESULTS ANALYSIS BY MULTIPLE

LINEAR REGRESSION
The general purpose of a multiple regression analysis is to establish a relationship (criterion or model)
652
Table 2. Numerical experimental design results for

masonry block parameters in the first developed model.
Test
(Number)
Cohesion
(MPa)
Tensile
strength
(MPa)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
1
1
1
1
1
1
1
1
1
0.5
0.5
0.5
0.5
0.5
0.5
0.5
0.5
0.5
0.2
0.2
0.2
0.2
0.2
0.2
0.2
0.2
0.2
1
1
1
0.5
0.5
0.5
0
0
0
1
1
1
0.5
0.5
0.5
0
0
0
1
1
1
0.5
0.5
0.5
0
0
0
Friction
angle
( )
Plastic
block
(model 1)
(Number)
30
15
0
30
15
0
30
15
0
30
15
0
30
15
0
30
15
0
30
15
0
30
15
0
30
15
0
1
9
24
1
9
24
16
22
55
29
76
112
29
77
112
46
96
116
116
116
116
116
116
116
116
116
116
Figure 9. Calculated regression factor coefficients and

their interactions (model 1).
The proposed linear model for the initial masonry

block shape thus becomes:
For the second shape of masonry blocks (model),

the linear model becomes:
between a dependent variable and one or more independent or predictor factors (Pedhazur, 1997). In our
case, the dependent variable is the number of plastic
masonry blocks PN while the exploratory variables
are: masonry block cohesion (C), masonry block tensile strength (Tr), and masonry block friction angle (),
respectively. The general form of the adopted linear
model including interaction terms is as follows:
Where PN is the dependent variable, 0 the constant and 1 , 2 , 3 , 13 , 23 and 123 the regression
coefficients of the various terms to be solved by the
regression technique.
The diagram in Figure 9 provides all of the calculated multiple regression factor coefficients, with
which a multiple regression was applied to the standardised experimental design factors. The standardisation process changes all factor values over the interval
[1, +1]. This multiple regression was performed
using the wellknown Mathematica software.
These two linear models reveal that among the three

studied factors, only cohesion (1 ) and friction angle
(3 ) significantly influence masonry block mechanical
behaviour, though the model does show that cohesion remains the single most significant factor of all
those studied. The interaction coefficients indicate that
only cohesion and friction angle exhibit a significant
interaction capable of influencing the response.
8
CONCLUSION
In this paper, an experimental design was proposed to

simulate ageing effects on old tunnel behaviour; a complete factorial experimental design, which expresses
the evolution of three selected masonry mechanical
properties, was then forwarded. The factors selected
for the present study were: masonry block cohesion,
tensile strength and friction angle.
All experimental design tests were modelled by
means of the distinct element method, executed with
the UDEC code.
Results analysis indicated that among the three
studied masonry block mechanical factors, only cohesion and friction angle along with their interaction
may exert a significant influence on masonry mechanical behaviour; furthermore, cohesion proved to be
653
the most significant of all factors studied. The noninfluence of tensile strength factor is explained by the
fact that the masonry structure is mainly loaded in
compression due to its vaulted section shape.
For a typical model of an old tunnel, only three
mechanical properties (factors) of masonry support
in old tunnels were studied. It is important to point
out however that several other mechanical properties
of masonry must be taken into account in order to
study their influence on the behaviour of older tunnel
supports. Our subsequent research goal will focus on
involving other masonry block and joint parameters in
studies on old tunnel behaviour.
REFERENCES
Barrentine, L.B. 1999. An Introduction to Design of Experiments:A SimplifiedApproach. Published by theAmerican
Society for Quality, Chapter 3: experiments with three
factors: 2735.
Brookes, C.L. & Mullett, P.J. 2004. Services load testing,
numerical simulation and strengthening of masonry arch
bridges, CIMNE, Barcelona: 10 pages.
Bicanic, N., Stirling, C. & Pearce, C.J. 1995. Discontinuous
modelling of masonry bridges, Computational Mechanics, V. 31: 293314.
Bicanic, N., Stirling, C. & Pearce, C.J. 2002. Discontinuous
Modelling of Structural Masonry. Fifth World Congress
on Computational Mechanics, Vienna, Austria: 18 pages.
CETu (Centre dEtude des Tunnels). 2004. Guide de
linspection du gnie civil des tunnels routiers, Ministre
de lquipement, des transports se du logement direction
des routes France. ISBN: 2-11-084749-2 : 95 pages.
Kerisel, J. 1975. Old structures in relation to soil conditions.
Geotechniques 25, No. 3, p. 433484.
Loureno, P.B. 2001. Analysis of historical constructions:
From thrust-lines to advanced simulations,Historical
Constructions P. Roca (Eds.) Guimare : 91116.
Ford, T.E., Augarde C.E. & Tuxford, S.S. 2003. Modelling masonry arch bridges using commercial finite element software, the 9th International Conference on Civil
and Structural Engineering Computing, Netherlands: 20
pages.
Giuriani, E., Gubana, A. & Arenghi, A. 2001. Structural

rehabilitation of masonry vault, Proceedings from the
UNESCO-ICOMOS Millennium Congress, Paris: 6 pages.
Hoek, E. 2000. Practical Rock engineering, Chapter 11, Rock
mass properties. rocscience.com on line: 161203.
Hughes, T.G. & Davies, A.W. 1998. The influence of soil and
masonry type on the strength of masonry arch bridges,
proc, sec, int . arch bridge conf., A. sinopoli Ed: 321330.
Idris, J., Verdel, T. & Al Heib, M. 2004 1. Feedback from
a database created for reporting accidents in tunnels and
galleries. Urban Transport X. Urban Transport and the
Environment in the 21st century. Dresden, Germany, WIT
press, ISBN 1-85312-716-7: 4150.
Idris, J., Al Heib, M. & Verdel, T. 2004 2. Base de donnes
des accidents et des incidents survenus dans les ouvrages
souterrains. Tunnels et ouvrages souterrains, No. 182:
363368.
Idris, J., Verdel, T. & Al Heib, M. 2007. Numerical modelling and mechanical behaviour analysis of ancient
tunnel masonry structures. Tunnelling and Underground
SpaceTechnology.(Article in press) :13 pages.
Janssen, H.J.M. 1997. Structural Masonry. Structural
Masonry, Numerical studies with UDEC. Centre for civil
engineering research and codes.A.A. Balkema, Roterdam,
Netherlands, ISBN 90 5410680 8: 96106.
Olofsson, T. 1985. A non-linear model for the mechanical
behaviour of continuous rock joints. In Proc. of International Symposium on Fundamental of Rock Joints, Ed
Stephansson O. Bjrkliden, Sweden: 395404
Pedhazur, E.J. 1997. Multiple regression in behavioral
research, third edition, Chapter 5, Multiple regression,
New York: Harcourt Brace College Publishers. ISBN
0-03-072831-2: 95135.
Sumon, S. K. & Ricketts, N. 1995. Strengthening of Masonry
Arch Bridges. Chapter in Arch Bridges. Publ Thomas
Tellord, London.
Valluzzi, M.R., Binda, L. & Modena, C. 2004. Mechanical behaviour of historic masonry structures strengthened
by bed joints structural repointing. Construction and
Building Materials V.19 : 6373.
Verdel, T. & Bigarre, P. 1999. Modlisation de tunnels
anciens avec le logiciel UDEC, Rapport INERIS, Socit
SIMECSOL : 112.
654
Excavation with stepped-twin retaining wall: Model tests and numerical

simulations
N. Iwata, H.M. Shahin, F. Zhang, T. Nakai, M. Niinomi & Y.D.S. Geraldni
Department of Civil Engineering, Nagoya Institute of Technology, Nagoya, Japan
ABSTRACT: Braced excavation using stepped-twin retaining wall is becoming popular in Japan. As it is a new
technique used to prevent movements of double-elevated ground, mechanism of deformation due to excavation
and change of stresses have not been fully understood. The design methodology of this technique is also not
properly established. In this research, two-dimensional model tests are conducted to investigate the deformation
mechanism of the ground and the earth pressure of the stepped-twin retaining wall. Numerical simulations with
finite element method are also carried out for the same scale of the model tests. The aim of the research is to make
clear the mechanism of the braced excavation using stepped-twin retaining wall and to establish an effective way
to evaluating the mechanical behaviors of the retaining wall and the surrounding ground.
INTRODUCTION
In urban area, open excavation often cause problems to

surrounding ground and adjacent structures. In practical daily design works, however, earth pressure is
usually predicted by conventional methods such as
a frame model together with Rankines earth pressure theory. There is also no appropriate method to
predict surface settlements of ground, which is usually predicted by empirical method and/or elastic
finite element method. For braced excavation using
stepped-twin retaining wall, conventional method cannot taken into consideration properly the influence of
nearby structures as well as the construction sequence
in evaluating the ground movements and the earth
pressure.
In this research, two-dimensional (2D) model tests
are conducted to investigate the deformation mechanism of the ground and the earth pressure of the
stepped-twin retaining wall. Numerical simulations
with finite element method are also carried out for
the same scale of the model tests. In the finite element analyses, subloading tij model (Nakai & Hinokio
2004), is used in the analysis to model the ground
material. This model can describe typical stress deformation and strength characteristics of soils such as the
influence of intermediate principal stress, the influence of stress path dependency of plastic flow and the
influence of density and/or confining pressure. Mass
of aluminum rods is used in the model ground. Several
patterns of the model tests are performed varying the
length of the retaining wall and changing the distance
Figure 1. 2D Model test device.
between the two walls. The results of stepped-twin

retaining wall are compared with the single retaining
wall.
2
OUTLINE OF MODEL TESTS AND

NUMRICAL SIMULATIONS
2D Model tests and the corresponding numerical

simulations of braced excavation using stepped-twin
retaining wall, were carried out to make clear the
mechanical behavior the problem. Figure 1 shows the
outline of the 2D Model test device. Four cases of
model tests with different length of outer retaining wall
and different spacing of stepped-twin retaining walls
were considered, as shown in Figure 3.
655
Table 1.
Material Parameters.
Ground
Aluminum rods
Unit weight = 20.4 (kN/m3 )
Aluminum plate
EI = 0.88 (kN*m2 /cm)
EA = 4.22 104 (kN/cm)
Upper: k1 = 3.64 (kN/m/cm)
Lower: k1 = 4.13 (kN/m/cm)
Retain wall
Strut
Table 2.
Parameters of ground made of aluminum rod mass.
N eNC at
p 98 kPa &
q = 0 kPa
Rcs = (1 /3 )cs
e
Figure 2. Outline of 2D Model test device.
Figure 3. Cases of study.
Figure 2 shows the schematic diagram of the twodimensional apparatus. The size of the model ground is
68 cm in width and 45 cm in height. Aluminum rods of
5 cm in length, having diameters of 1.6 mm and 3.0 mm
and mixed in the ratio of 3:2 in weight, are used as the
model ground (unit weight of the mass is 20.4 kN/m3 ).
In the experiment, the model ground was excavated
with a thickness of 1.5 cm each time and two struts,
located at the levels of 1.5 cm and 7.5 cm respectively, were set into place at the time when excavating
level reached 1.5 cm below its position. The retaining walls, with different length and spacing, were set
before the ground was excavated. Table 1 shows the
material parameters of the model ground, the retaining wall (Aluminum plate) and the struts. A laser type
0.008
0.004
0.3
1.8
0.2
1.2
1300
same parameters as
Cam-clay model
shape of yield surface (same

as original Cam-clay at = 1)
influence of density and
confining pressure
displacement transducer is used to measure surface

settlement of the ground. By taking photographs with a
digital camera and using image processing of the photos, the distribution of movement and consequently the
strain of the ground can be measured.
Numerical analyses are carried out with the same
scale of the model test considering plane strain condition using isoparametric element. An elastoplastic
constitutive model, named as subloading tij model
(Nakai & Hinokio, 2004) is used in the finite element analyses to simulate the mechanical behaviors of
the model ground. The model can take into consideration automatically the influence of the intermediate
principal stress, by introducing a modified stress tij
(Nakai and Mihara, 1984; Nakai and Matsuoka, 1986).
Subloading surface concept proposed by Hashiguchi
(1980) was also adopted in the model to consider
the influence of density of ground materials. Detailed
description about the performance and the reasoning of the model can be referred to aforementioned
references.
Table 2 lists the parameters of model ground made
of aluminum rod. Figure 4 shows the performance of
the model. Figure 5 shows the finite element mesh of
Case1-b. Smooth boundary condition is assumed for
side boundaries, and the bottom of the meshes is kept
fixed. The initial stresses of the ground are calculated
by simulating the self-weight consolidation by applying body forces, starting from a negligible confining
pressure (p0 = 9.8 106 kPa) and an initial void ratio
656
Figure 4. Stress-strain-dilatancy curves for aluminum rod

mass.
Figure 5. Finite element mesh (Case1-b).
e = 0.35. The retaining wall was simulated with beam

element and the strut is simulated with spring element.
Between the ground and the retaining wall, joint element whose mechanical behavior is simulated by a
perfect-plastic joint elements (Nakai, 1985), was introduced to consider possible sliding between the ground
and the wall.
Figure 6 shows the observed surface settlements at

different excavation stages in four cases. The length
of outer retaining wall does not affect too much the
settlement, while the spacing of twin walls has a great
influence on the settlement. The shorter the spacing is,
the larger the settlement will be. The same tendency
can be obtained in the correspondent calculations, as
shown in Figure 7.
Figures 8 and 9 show the observed and calculated
deflections of the retaining walls during the excavation. Similar to the surface settlement, the main factor
affecting the deflection is the spacing of twin walls.
From Figures 6 to 9, it is also known that the
numerical calculation conducted in this paper can not
only well describe the deformation patterns of the
ground and the retaining wall qualitatively but also
quantitatively to some extent.
Figure 6. Surface settlements (Observed).
In the calculation, the frictional angle of joint elements which are used to simulate the friction between
the ground and the wall is determined with constant
normal stress frictional test and is found to be 17
degree. In the calculation, however, the displacement
of the ground along the wall does not fit the observed
one well. This is due to the fact that a perfect-plastic
model is used for the joint elements which do not
allow any elastic deformation before the joint element
reaches yielding state. The reason why we do not use
elasto-plastic model is that it is difficult to determine
657
Figure 7. Surface settlements (Computed).
the shear stiffness of joint element. Further research

needs to be down in future.
Figures 10 and 11 show the displacements of the
ground surrounding the excavated area, obtained both
from model test and numerical calculation. The deformation of the ground is limited to the area near the
excavation. The calculated results agree well with the
observed ones.
Figures 12 and 13 show the distribution of shear
strain (The second invariant of strain tensor), obtained
both from model test and numerical calculation. It is
found that the shear strain of the ground is also limited
Figure 8. Displacements of retaining wall (Experiment).
658
Figure 9. Displacements of retaining wall (Computed).
Figure 10. Displacement

d = 180 mm).
of
ground
(Observed,
Figure 11. Displacement

d = 180 mm).
of
ground
(Computed,
to the area near the excavation. The calculated results

agree well with the observed ones.
Figure 14 gives a comparison between the observed
and the calculated results of axial forces within the
struts at different excavating stages. Similar to the
surface settlement, the main factor affecting the axial
force is the spacing of twin walls instead of the length
of the outer wall. The closer the spacing is, the higher
the axial force of the second strut will be. The first
659
Figure 12. Distribution of shear strain (Observed,

d = 180 mm).
Figure 14. Comparison of axial forces in struts.
Figure 13. Distribution of shear strain (Computed,

d = 180 mm).
strut, however, is not affected too much by these two

factors, that is, the length of wall and the spacing. The
numerical calculation can well describe the test results,
both qualitatively and quantitatively.
Figure 15 gives a comparison of calculated change
of earth pressures on retaining walls during excavation for the cases of 1-a and 1-b. It is found that the
earth pressures on inner retaining wall do not show
much evident difference.The outer wall, however, does
behave quite differently, that is, the passive earth pressure on the bottom in case 1-a (narrow spacing) is much
smaller than those in 1-b (wide spacing), implying that
the ground between the twin wall cannot be expected
to resist the deflection of the outer wall as what we
Figure 15. Change of earth pressures on retaining walls

during excavation.
usually expected to be a passive earth pressure when

the spacing of the two wall is enough narrow.
Figure 16 shows a comparison of calculated results
of axial forces in struts in different excavation methods. In single-wall excavation, the axial force in upper
strut increases firstly and then decreases after the
lower strut comes into action. In twin-wall excavation, however, both the axial forces in upper and lower
struts increase during the excavation. Meanwhile, the
660
pressure of the stepped-twin retaining wall. From this

research the following points can be concluded:
1. The displacements of the walls are inversely proportional to the distance between the walls.
2. The surface settlement follows the same tendency
of the wall displacements, and it is very much
dependent on the distance between the two walls.
3. The distance between the walls is more important
factor than the embedded length of the wall.
4. Unlike the single retaining wall the struts of the
stepped-twin retaining wall share axial load a more
efficient way.
Finite element analysis conducted in this paper,
which is based on subloading tij model and, is capable
to describe the mechanical behaviors of the twin-wall
excavation qualitatively and quantitatively.
REFERENCES
Figure 16. Comparison of different excavation methods.
axial force of lower strut increases much fast in twinwall excavation than those in single-wall excavation.
Therefore the mechanism of twin-wall excavation and
single-wall excavation is much different and should be
considered carefully in daily design.
4
CONCLUSIONS
Hashiguchi, K. 1980. Constitutive equation of elastoplastic materials with elasto-plastic transition. Jour. of Appli.
Mech. ASME 102(2): 266272.
Nakai, T. 1985. Finite element computations for active and
passive earth pressure problems of retaining problems.
Soils and Foundations 25(3): 98112.
Nakai, T. & Hinokio, M. 2004. A simple elastoplastic model
material parameters. Soils and Foundations 44(2): 5370.
Nakai, T. & Matsuoka, H. 1986. A generalized elastoplastic
constitutive model for clay in three-dimensional stresses.
Soils and Foundations 26(3): 8198.
Nakai, T. & Mihara, Y. 1984. A new mechanical quantity
for soils and its application to elastoplastic constitutive
models. Soils and Foundations 24(2): 8294.
Laboratory model tests and the corresponding numerical simulations are conducted for investigating the
deformation mechanism of the ground and earth
661
Stability of an underwater trench in marine clay under ocean wave impact

T. Kasper & P.G. Jackson
COWI A/S, Kongens Lyngby, Denmark
ABSTRACT: A long section of the immersed tunnel of the Busan-Geoje Fixed Link in South Korea is to be
constructed in a trench in soft marine clay. The site is exposed to large typhoon waves and the trench is to be
left open for approximately one year. The trench profile was chosen as a balance between excavated volume
(costs) and risk of slope instability under large waves. To evaluate this risk, far-shore wave data have been
transformed into near-shore wave conditions by means of numerical modelling and wave flume tests have been
carried. Potentially critical waves have been identified and their impact has been analysed by means of coupled
hydro-mechanical numerical simulations. Based on the strength reduction method, safety factors for the slope
stability during wave impact have been determined. The results and the consequences for the economic design
of such a trench subject to large waves are discussed.
INTRODUCTION
The Busan-Geoje Fixed Link between South Koreas

second largest city Busan and the island of Geoje consists of a 3.2 km long immersed tunnel (Odgaard et al.
2006), two rock tunnels and two 1.7 and 1.9 km long
cable-stayed bridges. About 2.2 km of the immersed
tunnel is constructed in a 12 to 15 m deep trench in
soft marine clay at water depths between 20 and 50 m.
The area is prone to frequent raids of typhoons and the
trench is dredged approximately 1 year before construction of the tunnel. When the tunnel is placed
in the trench, the lower half of the trench is filled
with tunnel protection material (backfill). It was therefore necessary to design the trench profile as a balance between the excavated volume, i.e. construction
costs on the one hand and the risk of slope failure
under wave impact within a 1 years period on the
other hand.
This paper discusses in Section 2, how the wave
characteristics at the site have been determined and
how the most critical waves with regard to trench slope
stability have been identified. Based on the geotechnical characterization of the soft marine clay presented in
Section 3, the impact of the critical waves on the trench
is investigated by means of coupled hydro-mechanical
numerical simulations (Section 4). Strength reduction
analyses are made in the simulations in regular intervals of one second to determine the failure modes and
factors of safety (Dawson et al. 1999). The last part
of the paper contains a discussion of the results and
general conclusions.
Figure 1. Numerical wave modelling: Calculated significant wave heights for a 10,000 years return period. The tunnel
is indicated by the dashed line.
DERIVATION OF DESIGN WAVES
2.1 Numerical wave modelling

Based on a statistical analysis of wave and wind data
in the area, the extreme far-shore boundary waves and
extreme wind speeds have been derived. From these
input data, the wave conditions at the location of the
tunnel have been determined by means of numerical wave modelling with the program MIKE 21 (DHI
Water & Environment).
The obtained maximum significant wave heights
along the tunnel alignment are 6.2 m for a return period
of 10 years, 8.0 m for a return period of 100 years and
9.2 m for the 10,000 years wave event (Figure 1). The
large waves come from southerly direction with a wave
velocity of approximately 13 m/s and pass the tunnel
663
Figure 2. Effect of typhoon waves on a breakwater in Busan

(typhoon Maemi, 12.9.2003): (a) before and (b) after the
storm.
Figure 4. Represenation of undrained strength in the numerical model with the Mohr-Coulomb model based on effective
strength parameters.
Marine clay
cu (kPa) in situ
10 years wave
100 years wave
-2
-4
-100
0.0
10,000 years wave

-50
50
Depth (m)
Excess pressure (m)
Wave pressure profiles at seabed
100
Length (m)
Figure 3. Most critical seabed pressure profiles for the

trench stability determined from the wave flume tests.
0.0
2.0
4.0
6.0
8.0
10.0
12.0
14.0
16.0
10.0
20.0
30.0
40.0
Measured lower
bound (SHANSEP,
prec. stress 10 kPa
Measured upper
bound (SHANSEP,
prec. stress 65 kPa)
Numerical model
(Mohr-Coulomb)
Figure 5. Comparison between measured and modelled

undrained strength of the marine clay for in situ conditions.
trench almost perpendicularly. The potential devastating effect of such large waves is illustrated for a
breakwater in Figure 2.
2.2 Wave flume model tests
Based on the results of the numerical wave modelling,
wave flume tests have been carried out for tunnel element TE7, which is subject to the highest waves at a
relatively shallow water depth of 20 m. Each test consisted of a series of several hundreds of waves. From
the pressure measurements in the tests, the seabed
pressure profiles with the largest pressure difference
across the 32 m wide trench slopes have been identified. These pressure profiles shown in Figure 3 are
considered to be most critical for the slope stability. It
should be noted that due to hydrodynamics, the seabed
pressures under waves do not simply correspond to the
physical wave height above.
3
GEOTECHNICAL PROPERTIES
Extensive ground investigations and laboratory tests

have been carried out for the project (Steenfelt et al.
2008). The undrained strength of the clay has been

assessed from 30 cone penetration tests along the
alignment and can be described according to the
SHANSEP approach (Steenfelt & Foged 1992) as
with the preconsolidation stress of the marine clay

typically ranging between 10 and 30, max. 65 kPa.
The effective strength parameters = 25 , c = 3 kPa
have been derived from 19 consolidated undrained
triaxial tests. In the numerical model, the material
strength is described with the Mohr-Coulomb model.
Due to the fact that the clay is only slightly overconsolidated, the Mohr-Coulomb model can also be used
without modification to correctly model the undrained
strength in case of undrained conditions (Figure 4 and
Figure 5).
The stiffness properties and permeabilities have
been derived from oedometer tests. The material
parameters are summarised in Table 1.
664
Material parameters for the numerical simulations.
Material model
sat (kN/m3 )
K0 (-)
E (kPa)
(-)
( )
c (kPa)
k (m/s)
JOB TITLE :
Marine clay
Alluvium
Mohr-Coulomb
14.7
0.6
8000
0.2
25
3
1 109
Mohr-Coulomb
20
0.426
50000
0.25
35
0
1 105
Excess pressure (m)
Table 1.
2
10 years wave
Approximation
-2
-4
-100
-50
50
100
Length (m)
Figure 7. Approximation of the 10 years wave pressure profile in the FLAC model (A1 = 3.4 m, L1 = 70 m, A2 = 3.4 m,
L2 = 80 m).
Trench stability analysis
hydraulic and mechanical boundary conditions along

the seabed and the trench surfaces. The wave pressure
profile is described by means of two sine curves, one
for the wave crest and one for the wave trough. This
allows to consider different shapes of the wave crest
and wave trough according to
FLAC (Version 5.00)
100
LEGEND
12-Mar-07 23:19
step 149210
Flow Time
0.0000E+00
-1.089E+01 <x< 2.069E+02
-9.239E+01 <y< 1.254E+02
60
User-defined Groups
Alluvium
Marine clay
20
Grid plot
0
Wave pressure profile at seabed

4
5E+01
-20
-60
COWI A/S
Denmark
20
60
100
140
180
Figure 6. FLAC model of the trench.
4
4.1
NUMERICAL SIMULATION OF WAVE

IMPACT
Model description
The effect of waves on the trench is modelled by means

of numerical simulations using the finite difference
program FLAC (Itasca 2005). The two-dimensional,
plane-strain model is shown in Figure 6.
The vertical boundaries of the model are fixed in
horizontal direction, while the bottom of the model
is fixed both in horizontal and vertical direction. The
vertical boundaries and the bottom are assumed to be
impermeable.
Soil stresses and pore water pressures are initialised
according to the self-weight of the soils, K0 = 1
sin and a still water table at + 63 m model coordinate
(still water pressures pstill ).
In a first step, the trench excavation is modelled
either as a drained or undrained process. This is done
to allow for modelling of wave impact shortly after
trench excavation (undrained excavation modelling)
and wave impact 1 year after trench excavation, where
the excess pore pressures in the clay in the relevant
area next to the trench slopes will almost completely
be dissipated (drained excavation modelling).
Afterwards, wave impact is modelled by means
of coupled hydro-mechanical analyses with transient
The described approach is able to model the wave

pressure profiles quite accurately, as illustrated for the
10 years wave in Figure 7.
Based on Eqs. (2) and (3), the wave pressures are
evaluated for each point in time at each gridpoint
along the seabed and the trench surfaces and the corresponding total normal stress and pore water pressure
boundary conditions
are applied. It should be noted that water pressures

are defined positive, while compressive stresses are
defined negative. According to the principle of effective stresses, these boundary conditions imply that the
effective normal stresses at the seabed and the trench
surfaces remain 0. Based on a linear variation of the
boundary conditions along each zone edge, the whole
wave pressure profile is approximated as a piecewise
linear function with the correct values at each gridpoint. The model is defined such that a wave train
with an arbitrary number of waves can be simulated.
As a compromise between computational effort and
concern about a possible change in safeties from wave
to wave, a wave train with 2 waves is considered in the
basic simulations. Due to the fact that each of the investigated large waves passes the trench within a period
of approximately 6 seconds, the simplified assumption
of quasi-static conditions in the simulation appears to
be adequate. The compressibility of the pore water is
665
JOB TITLE :
Table 2.
Comparison of safety factors without wave impact.
FLAC (Version 5.00)
100
LEGEND
12-Mar-07 23:19
step 149210
Flow Time
0.0000E+00
-1.089E+01 <x< 2.069E+02
-9.239E+01 <y< 1.254E+02
Undrained excavation
Drained excavation
60
Boundary plot
0
5E+01
20
Factor of Safety 1.86

Max. shear strain-rate
0.00E+00
1.50E-07
3.00E-07
4.50E-07
6.00E-07
7.50E-07
9.00E-07
1
2
FLAC
GEO-SLOPE
PLAXIS
2.67
1.86
2.622
1.831 , 1.832
2.64
1.87
General limit equilibrium solution

Solution based on finite element stresses
-20
Contour interval= 1.50E-07
-60
COWI A/S
Denmark
60
100
140
180
Figure 8. Predicted failure mechanism without wave impact

after full consolidation of the trench excavation (drained
conditions): Safety factor 1.86.
taken into account by a bulk modulus Kw = 2 109

Pa (pure water), while the compressibility of the soil
grains is neglected. Failure mechanisms and safety factors are determined in time intervals of 1 second during
wave impact by means of strength reduction analyses
(Dawson et al. 1999).
4.2
Figure 9. Illustration of results for drained conditions:

(a) Traditional limit equilibrium solution (SLOPE/W),
(b) Finite element strength reduction solution (PLAXIS).
Safety factors during wave impact after full
consolidation of the excavation
4.50
4.00
Safety factor (-)
20
Results
Figure 8 shows the result of a safety analysis without

wave impact after full consolidation of the trench excavation. The failure mechanism is illustrated by means
of maximum shear strain rates at the end of the strength
reduction analysis.
The FLAC result can be verified by comparison
with other solutions (Table 2 and Figure 9). The
drained results can be compared with a traditional
limit equilibrium solution and a finite element stress
based solution obtained with the program SLOPE/W
(GEO-SLOPE) as well as with a strength reduction
analysis with the finite element program PLAXIS
(Brinkgreve & Bakker 1991). In the case of undrained
conditions, a FE calculation is needed with all programs in order to correctly predict the effective stresses
and excess pore pressures. The shapes of the slip
surfaces and the safety factors show satisfying agreement in all cases. It is obvious that the safety factors
decrease with consolidation. The reason is that the
excavation represents an unloading process leading
to a time-dependent reduction in effective vertical
stresses, which are crucial for the soil strength.According to the Mohr-Coulomb failure criterion the shear
strength is linearly dependent on the effective normal
stress. It is estimated that after one year, the conditions
in the clay in the critical area close to the slope surfaces
may be nearly drained.
Figure 10 shows the evolution of the computed
safety factors during wave impact after full consolidation of the trench excavation. It can be observed that
3.50
3.00
10 years wave
2.50
100 years wave
2.00
10,000 years wave
1.50
1.00
0.50
0
10
20
30
40
50
Time (s)
Figure 10. Computed safety factors during wave impact

after full consolidation of the trench excavation.
the safety factors show a large variation in the range

between 0.91 and 4.12 during passage of the waves.
The safety factors exceed 1.86 obtained for still
water when a wave crest is above the trench. Then,
increased pressures act along the bottom of the trench
and decrease along the trench slopes towards the crown
points of the trench (Figure 11). This pressure distribution in the trench counteracts the destabilizing gravity
forces, thus leading to increased safety factors of up
to 4.12.
On the other hand, a wave trough above the trench
causes high pressures on the seabed beside the trench
which decrease down along the trench slopes (Figure 12). This pressure distribution adds to the gravity
forces, thus leading to decreased safety factors of down
to 0.91. The prediction of safety factors smaller than
1 for the 100 and 10,000 years wave has been facilitated by scaling the clay strength (tan and c ) up and
scaling the obtained safety factors down.
According to Table 3, the difference in minimum
safety factors between the 10, 100 and 10,000 years
666
JOB TITLE :
FLAC (Version 5.00)
100
LEGEND
13-Mar-07 5:25
step 109770
Flow Time
1.6000E+01
-1.089E+01 <x< 2.069E+02
-9.239E+01 <y< 1.254E+02
60
Boundary plot
0
5E+01
20

0.00E+00
2.00E-07
4.00E-07
6.00E-07
8.00E-07
1.00E-06
1.20E-06
-20
-60
COWI A/S
Denmark
20
60
100
140
180
Figure 11. 10 years wave, t = 16 s: Illustration of wave pressure profile and predicted failure mechanism with the largest
safety factor of 4.12 for wave impact after full consolidation
of the trench excavation.
JOB TITLE :
FLAC (Version 5.00)
100
LEGEND
13-Mar-07 23:11
step 203328
Flow Time
2.2000E+01
-1.089E+01 <x< 2.069E+02
-9.239E+01 <y< 1.254E+02
60
Boundary plot
0
5E+01
20

0.00E+00
1.50E-08
3.00E-08
4.50E-08
6.00E-08
7.50E-08
9.00E-08
-20
-60
COWI A/S
Denmark
20
60
100
140
180
Figure 12. 10 years wave, t = 22 s: Illustration of wave

pressure profile and predicted failure mechanism with the
smallest safety factor of 1.01 for wave impact after full
consolidation of the trench excavation.
Table 3.
Summary of minimum safety factors.
10 years wave
100 years wave
10,000 years wave
Wave impact
shortly
after trench
excavation
Wave impact after

full consolidation
of the trench
excavation
1.38
1.29
1.24
1.01
0.95
0.91
modelling of the excavation). Graphs with the same

general pattern, but with larger safety factors are
obtained for wave impact shortly after excavation of
the trench (undrained modelling of the excavation).
It may be concluded from the calculated safety
factors that failure of the slopes may occur if these
extreme waves would pass the trench after 1 year
consolidation. However, it can be argued that sliding of slopes is a dynamic process which takes place
in the range of at least several seconds. Figure 11
and Figure 12 illustrate that the maximum and minimum safety factors are only a few seconds apart.
Even if the exceptional 100 and 10,000 years waves
would pass the open trench after 1 year consolidation, larger sliding movements may be prevented by
the fact that critical safety factors occur only within
about 1 or 2 seconds. Although water wave impact is
a slower phenomenon, the situation may be compared
with earthquake impact where the exceedance of yield
accelerations in shorter periods induces some irreversible plastic deformations which do not necessarily
lead to complete failure. An estimate of expected sliding movements during wave impact could be obtained
e.g. by means of dynamic coupled hydro-mechanical
simulations with the dynamic version of FLAC.
A simulation of the 10 years wave event with
Kw = 2 107 Pa (representing pore water with a high
content of dissolved/entrapped air) yields a minimum
safety factor of 0.99 for wave impact after full consolidation of the excavation and 1.35 for wave impact
shortly after excavation of the trench. These safeties
for highly compressible pore water are slightly smaller
than the corresponding safeties of 1.01 and 1.38 for the
basic case. The influence of pore water compressibility on the behaviour of different soil structures under
draw down and wave impact has been investigated in
detail e.g. by Khler (2000).
No perceptible change in safety factors can be
observed between the first and second wave in the simulations (Figure 10). This has further been confirmed
by a simulation with a wave train of 6 waves.
waves is relatively small and the safety factors for wave

impact after full consolidation of the trench excavation
are approximately 36% lower than the corresponding safety factors for wave impact shortly after trench
excavation.
Figure 10 shows the safety factors for wave impact
after full consolidation of the excavation (drained
DISCUSSION AND CONCLUSIONS
Due to frequent raids of typhoons in the area, it was

necessary to consider the possible effect of large waves
in the design of the 2.2 km long and 12 to 15 m deep
trench for the immersed tunnel of the Busan Geoje
Fixed Link in South Korea. Safety factors for the
trench stability under extreme waves with return periods of 10, 100 and 10,000 years have been determined
by means of numerical simulations. The minimum
values have been found to be between 1.24 (10,000
years return period) and 1.38 (10 years return period)
for wave impact shortly after trench excavation and
between 0.91 (10,000 years return period) and 1.01
667
(10 years return period) for wave impact at the end

of the construction period (after consolidation of the
trench excavation). These safety factors have been
considered to be acceptable for the following reasons:
The investigated situation is a temporary construction phase, where the trench is open for about 1 year.
Under these circumstances, the 10 years wave event is
considered to be the relevant design case.
Although the minimum safety factors for the 100
and 10,000 years waves are slightly below 1 (Table 3),
failure is unlikely even under these extreme waves due
to the fact that the minimum safety factors occur within
only 1 or 2 seconds (Figure 10). This period is considered to be too short for complete failure to occur. High
strain rates would be required for failure to develop
and it is a well-established fact that the strength of
clay is actually higher for high strain rates. Minimum
and maximum safety factors are only a few seconds
apart (Figure 10).
Failure of the trench slopes represents a financial risk and would require a clean-up afterwards, but
would neither affect lifes nor the technical success of
the project. The costs for excavation of a trench in offshore conditions at water depths between 20 and 50 m
are high and should be limited to a minimum.
6
PROJECT PROGRESS
The excavation of the trench began in July 2006 and

was finished in September 2006. Busan only experienced a minor tropical storm in summer 2006. The
production of the soil improvement (cement deep mixing) of the marine clay in the trench for the foundation
of the tunnel began in November 2006 and has been
finished in May 2007. The offshore construction of
the tunnel is scheduled to start at the end of 2007 and
finish in 2010.
REFERENCES
Brinkgreve, R.B.J. & Bakker, H.L. 1991. Non-linear
finite element analysis of safety factors. In G. Beer,
J.R. Booker & J.P. Carter (eds), Computer Methods
and Advances in Geomechanics: 11171122. Rotterdam:
Balkema.
Dawson, E.M., Roth, W.H. & Drescher, A. 1999. Slope
stability analysis by strength reduction. Gotechnique
49: 835840.
DHI Water & Environment, Denmark. MIKE 21. www.
dhisoftware.com/mike21/
GEO-SLOPE International, Ltd. GeoStudio 2004, version
6.19. www.geo-slope.com.
Itasca Consulting Group, Inc. 2005. FLAC Fast Lagrangian
Analysis of Continua, version 5.0. www.itascacg.com.
Khler, H.J. 2000. Pressure spreading at soil water interfaces and its influence on soil structure design. In
A. Cancelli et al. (eds), EUROGEO2000; Proc. Second European Geosynthetics Conference, Bologna, 1518
October 2000: Vol. 2, 687694. Bologna: Patron Editore.
Odgaard, S.S., Jensen, O.P., Kasper, T., Yoon, Y.H.,
Chang, Y., & Park, R.Y. 2006. Design of long immersed
tunnel for highway in offshore conditions Busan
Geoje Fixed Link. Tunnelling and Underground Space
Technology 21(34), Special Issue: Safety in Underground
Space (CD-ROM Proceedings of the ITA-AITES 2006
World Tunnel Congress, Seoul, Korea).
PLAXIS b.v. PLAXIS version 8.5. www.plaxis.nl.
Steenfelt, J.S. & Foged, N. 1992. Clay till strength
SHANSEP and CSSM. NGM-92, 8186.
Steenfelt, J.S., Jackson, P.G., Christensen, C.T., Lee, J-S., &
Ha, Y.-B. 2008. Ground investigations for the Busan
Geoje Immersed Tunnel. Submitted to 3rd International
Conference on Site Characterisation, ISC3, Taipei, 2008.
Zienkiewicz, O.C., Humpheson, C. & Lewis, R.W. 1975.
Associated and non-associated visco-plasticity and plasticity in soil mechanics. Gotechnique 25(4): 671689.
ACKNOWLEDGEMENTS
The authors gratefully acknowledge the permission
from Daewoo Engineering & Construction to publish
this paper.
668
A study on behavior of 2-arch tunnel by a large model experiment

S.D. Lee
Department of Civil Engineering, Ajou University, Korea
K.H. Jeong, J.W. Yang & J.H. Choi

Dodam Engineering and Construction Co., Korea
ABSTRACT: It is tendency that the parallel tunnels are constructed close to each other in order to diminish
civil complaints and environmental damage. The 2-Arch tunnel is similar to two parallel tunnels with very
short centre-to-centre distance. Recently, construction of 2-Arch tunnel is increasing. However, it is executing
without enough studies for behavior of the 2-Arch tunnel. In this research, a study for behavior of 2-Arch
tunnel is examined using large model test machine. At first we make the model ground with horizontal joint-set.
Then embody in-situ stress by applying pressure to boundary of model ground. Then excavate model ground
according to construction steps of the 2-Arch tunnel. As a result, during excavation of pilot tunnel, measured
ground displacements are about 4050% of whole displacement, which is concentrated in 0.25D (D: diameter
of tunnel) region around tunnel. Height of loosen area by construction of the 2-Arch tunnel is 0.15w (w:centreto-centre distance between left and right tunnel). These results are compared with DEM to conform the reliability
of results.
INTRODUCTION
The 2-Arch tunnel is similar to two parallel tunnels

with very short centre-to-centre distance. The 2-Arch
tunnel has been increasing in order to diminish environmental damage and ensures linking with structure
close by tunnel. There have been a series of researches
on the behavior of 2-Arch tunnel in soil and weathered
rock. However, the use of 2-Arch tunnel in korea has
been rapidly increasing in hard rock. So, researches
of 2-Arch tunnel in hard rock are urgently needed.
The stability of center upper part of 2-Arch tunnel is weak comparing with other parallel tunnels;
therefore it is necessary to install pillar and secure
structural stability. In 2-Arch tunnel, mutual effect of
precedence and afterward tunnel is bigger than general parallel tunnel. The stability of pillar and mutual
effect are very important in research of the 2-Arch
tunnel.
Therefore, this study is conducted to model characteristic of sedimentary rock that similar with ground
condition of prototype tunnel. Step in different stage
displacement tendency, precedence tunnel stress transition by excavate afterward tunnel, pillar and lining
behavior examine by achieving an experiment according to carrying out construction steps of 2-Arch tunnel.
With this, compare with numerical analysis (DEM) in
same condition and confirm the reliability of results.
SUMMARY OF EXPERIMENT
2.1 The model package

The dimensions of the model ground were 3 m 3 m
0.27 m (B H T). The main characteristic of the test
machine is application of both the vertical and lateral
loading system to embody in-situ stress.
Figure 1. Large model test machine.
669
Table 1.
Property of model ground.
Block
Joint surface
E
(MPa)
(KN/m3 )
C
(MPa)
( )
C
(MPa)
( )
980
0.25
19.8
2.55
35
32
Figure 2. The exist 2-Arch tunnel cross-section.
The inside of the model container was coated with

grease to reduce side friction so that shear stress transfer at the ground and the container interface could be
minimized.
2.2 The model ground
It has been well known that rock mass is not homogeneous and exists discontinuities. The behaviour of
tunnels and surrounding ground are heavily dependent
on the characteristic of the rock mass. In this study to
model characteristic of sedimentary rock, the model
ground was made of a number of concrete bricks. The
model ground was classified by Rock Mass Rating
(RMR) used in tunnel design in Korea widely. The
model ground used in the model tests had slightly
rough surface with aperture thickness less than 1 mm.
Overall the model ground had total RMR rating of 67
and hence the model ground was classified as class
number II, i.e., Good rock.
Figure 3. Improved 2-Arch tunnel cross-section.

Table 2. Application of reduction ratio.
Tunnel
Prototype
tunnel
Model
tunnel
E(ground)
(MPa)
E(lining)
(MPa)
Diameter Thickness
(m)
lining (cm)
9.8*103
1.96*105
23
40
1.2
0.6
9.8*10
1.96*10
2.3 Tunnel cross section and stiffness

reduction rate
There is a problem in problem in passing of equipment and construct of pillar, because center tunnel is
small and narrow (Fig. 2). Therefore, an experiment
achieved into improved 2-Arch tunnel section that considers equipment exit and entrance center tunnel of
approximately enlarge, and improve waterproof sheet
establishment location and carrying out order (Fig. 3).
Usually, the property of the material and size are
decided by through the suitable stiffness reduction rate
when achieve a model experiment. The property of
model tunnel was decided by the method of Duddeck
and Erdmann (1985) that use stiffness ratio () of lining. Actuality stiffness ratio of research object tunnel
and model tunnel depended on stiffness of each ground
and property of tunnel lining. Those are expressed with
below way. The reduction rate applications are same
with table 2.
Stiffness ratio of prototype tunnel
: = (Ek R3 )/(Eb Ib )
Stiffness ratio of model tunnel
: = (Ekm R3m )/(Ebm Ibm )
Ek (Ekm ): Modulus of elasticity of ground (MPa)
R(Rm ): Diameter of prototype(model) tunnel (cm)
Figure 4. Measure point of ground displacement.
Eb (Ebm ): Modulus of elasticity of lining (MPa)

Ib (Ibm ): Second moment of area lining
t(tm ): Thickness of prototype(model) tunnel lining
(cm)
2.4 Instrumentation
In the model test the behaviour of the tunnel and the
ground due to excavation was monitored using LVDT
(Linear Variable Displacement Transducer, ground
displacements) and load cell (pillar load). The locations of LVDTs are shown in Figure 4.
670
Thereafter the pillar load has been arranged to zero

to measure rock load, exclusively associated with the
main tunnel excavation.
2.6
Experiment condition and method
The model test was conducted under a uniform surface

loading of 980 kPa. An achieved experiment considers
actual construction step as follows.
1. Made up horizontality being stratiform rock mass
by the use concrete block of fixed size.
2. The model test was conducted under a uniform
surface loading of 980 kPa with lateral earth pressure of 1470 kPa and hence the earth pressure
coefficient was 1.5.
3. Excavation of upper part of the pilot tunnel.
4. Excavation of lower part of the pilot tunnel.
5. Arrangement of load cell reading to zero.
6. Excavation of precedence tunnel.
7. Excavation of afterward tunnel.
3 THE RESULTS OF THE EXPERIMENT
This research is that investigate behavior of 2-Arch
tunnel that is conducted through a large model test
and DEM analysis from the stratified rock. The main
examination contents are same as follows.
1. Displacement tendency at different steps
2. Precedence tunnel stress transition by afterward
tunnel excavation
3. Estimate of behavior of center pillar and loosen area
rock load.
3.1
Displacement tendency of different steps
The ground displacement is the main factor for tunnel

and adjacent structures stability judgment by tunnel
excavation. The displacements that appear in experiment results are compared with numerical analysis
(DEM) in equal condition and analyzed. Boundary
condition of this numerical analysis set limits to horizontality displacement in side wall and lower part
set limits to perpendicular displacement. Input data
of numerical analysis is decided by axial compression
test, direct shear test and RMR value of model ground.
Figure 5. The test procedures.
2.5 The center pillar

In order to measure the rock load acting on the centre
pillar, load cell was installed prior to the formation
of the model ground. In the real tunnel excavation,
the centre pillar is to be installed after completion of
the pilot tunnel. However, to measure loosening load
during the pilot tunnel excavation the centre pillar was
installed prior to the assembly of the model ground.
3.1.1 Displacement concentration region

In the model test, most of the ground displacement
was examined by large model test and numerical analysis occurred within 0.25D. Experiment consequences
and numerical analysis consequences of displacement
were variance more or less.
This is judged by joint action that is happen vertically joint between block and block that can produce
cause and experiment special quality upper that did
671
Table 3.
H-displacement of different steps (mm).

Horizontal displacement (mm)
Measure
point
Large
model
test
Horizontally
displacement (mm)
UDEC
8.0
7.0
6.0
5.0
4.0
3.0
2.0
1.0
0.0
2
4
29
30
2
4
29
30
1.49
1.36
1.4
1.14
1.943
1.423
1.617
1.231
1.83
1.62
1.54
1.33
3.578
2.254
2.344
1.618
1.85
1.69
1.55
1.33
3.577
2.254
2.344
1.618
3.21
2.18
2.02
1.87
7.127
5.203
4.984
3.280
3.38
2.31
2.17
1.94
7.349
5.382
5.141
3.387
initial
M-T No.2
M-T No.4
M-T No.29
M-T No.30
UDEC No.2
UDEC No.4
UDEC No.29
UDEC No.30
(a) H-displacement of different steps
Figure 6. Displacement concentration region.
Horizontally
displacement (mm)
8.0
not accord with correctly properties value of actuality

model ground in numerical analysis.
But, a model experiment and numerical analysis
result of displacement concentration extent and tendency are similar, therefore in identical branch result
was judged that it is meaning that compare mutual and
analyzed displacement to different steps.
7.0
modeltest
6.0
UDEC
5.0
4.0
3.0
2.0
1.0
0.0
point 2
point 4
point 29
point 30
(b) Completion of tunnel excavation
3.1.2 Horizontal displacement of different steps

In order to evaluate stability of the side wall, the lateral
ground displacements at the right part of the tunnel
have been monitored using LVDTs. The direction of
displacement is positive to tunnel and displacement of
opposite direction is negative.
a.
b.
c.
d.
e.
Excavation upper part of the pilot tunnel.

Excavation lower part of the pilot tunnel.
Installation of pillar.
Excavation of the precedence tunnel.
Excavation of the afterward tunnel.
During excavate to pilot tunnel, most of displacement occurred. In general the stress condition of the
ground after tunnel excavation may be different from
the in-situ condition. In particular, direction and magnitude of the principle stresses will be changed due to
stress re-distribution through the arching effect, hence
the measured results are indicate that majority of the
Figure 7. H-displacement.
stress change occurred during stage b after the stress

condition may not be changed much. Also, The measured results show that the longer the distance from
the tunnel base the smaller the lateral ground displacement.(measure point 2,4s horizontal displacement >
measure point 29,30s horizontal displacement). Overall the measurements suggest that the stability of the
2-arch tunnel depends mainly on the excavation of
the pilot tunnel. Therefore, the pilot tunnel should
be stabilized prior to excavation of the other part of
the tunnel. A large model experiment and numerical
analysis result is same with table 3.
3.1.3 Vertical displacement of different steps
In order to measure vertical ground displacements during the tunnel excavation at different steps, ground
672
Table 4. V-displacement of different steps (mm) M.T.
initial
Measure
point
7
13
8
14
9
15
10
16
1.05
0.82
1.24
0.99
1.32
0.99
1.2
0.86
1.53
1.21
1.66
1.25
1.72
1.31
1.59
1.24
1.55
1.16
1.67
1.35
1.74
1.39
1.59
1.3
1.84
1.38
2.01
1.59
2.52
1.94
1.94
1.51
2.33
1.89
2.99
2.38
2.74
2.11
2.21
1.81
0.25D
0.50D
0.25D
0.50D
0.25D
0.50D
0.25D
0.50D
vertical displacement (mm)
0
-1
-2
-3
-4
-5
-6
-7
UDEC No.7
UDEC No.8
UDEC No.9
UDEC No.10
M-T No.7
M-T No.8
M-T No.9
M-T No.10
(a) 0.25D from upper part of tunnel

Table 5. V-displacement of different steps (mm) UDEC.
initial
Measure
point
7
13
8
14
9
15
10
16
0.95
0.95
1.43
1.29
1.8
1.43
0.95
0.97
1.44
1.63
2.01
2.05
2.43
2.22
1.44
1.65
1.45
1.63
2.01
2.05
2.43
2.22
1.44
1.66
1.75
1.95
2.60
2.67
3.99
3.66
3.26
3.09
4.39
4.26
5.89
5.41
6.14
5.63
4.09
4.05
0.25D
0.50D
0.25D
0.50D
0.25D
0.50D
0.25D
0.50D
0
-1
-2
-3
-4
-5
-6
M-T No.13
M-T No.14
M-T No.15
M-T No.16
UDEC No.13
UDEC No.14
UDEC No.15
UDEC No.16
(b) 0.5D from upper part of tunnel

point 7,13
point 8,14
point 9,15
point 10,16
Figure 9. V-displacement of different stages.
-1
0.25D extent. Such result means that stability of whole

tunnel is dominated by stability of pilot tunnel excavation in 2-Arch tunnel. This suggests that rock bolt
length should be longer than 0.25D to prevent the rock
loosening.
-2
-3
-4
-5
-6
-7
modeltest 0.25D
modeltest 0.50D
UDEC 0.25D
UDEC 0.50D
Figure 8. V-displacement (completion of excavation).
3.2 Precedence tunnel stress transition by

afterward tunnel excavation
displacements at 0.25D and 0.5D above the tunnel

crown have been monitored (0.25D: points 710 0.5D:
points 1316), where D is the width of the tunnel
(Table 4).
The majority of the displacements occurred at excavation of pilot tunnel after insignificant changes of
the ground displacement had developed. The Vertical displacement in 0.25D was observed greatly 120%
when compare with displacement in 0.5D and most
displacement was concentrated on 0.25D extent.
Horizontal displacements more than 40% of whole
displacements and vertical displacements more than
20% are occurred during excavation of pilot tunnel. The displacement of model test is larger than
numerical analysis result (Table 5) at pilot tunnel excavation steps. The displacement was concentrated on
Precedence tunnel stability by afterward tunnel excavation is main concerns to establishment 2-arch tunnel
that parallel tunnel is very near. In the case of precedence tunnel, displacement is converged at completion of excavation and establishment of support. But
ground displacement and additional loading to support is increased by excavation of afterward tunnel.
Additional load that happen to precedence tunnel at
afterward tunnel excavate is 1030% of whole load
to the tunnel. (Lee and kim 2001, The consideration of improve 2-Arch tunnel Design and construction
method).
At the large model test result, displacement of
precedence tunnel is occurred 812% of whole displacement by excavation of afterward tunnel. And
2035% of UDEC analysis result increased at the
same point. It is considered similar result if take into
673
Vertical displacement (mm)
7.0
6.0
5.0
4.09
3.99
4.0
3.0
6.14
excavation completion of
precedence tunnel
excavation completion of
afterward tunnel
3.26
2.52
2.74
1.94
2.0
2.21
1.0
0.0
M-T No.9
M-T No.10
UDEC No.9
UDEC No.10
Figure 10. Displacement of precedence tunnel by excavation of afterward tunnel.
Figure 11. The load acting on pillar of 2-Arch tunnel in soil

and weathered rocks (Matsuda, 1998).
40.0
3.3
Behavior of the center pillar
Matsuda (1998) proposed an empirical equation of

rock load for a tunnel in soils and weathered rocks
by considering soil depth (H) and tunnel centre-tocentre distance (B). However, to date studies of rock
load that can be used in preliminary tunnel design for
2-arch tunnel in rock is rather limited. In the current
study, an empirical relation for rock load based on the
measurement is proposed for the rock when RMR is
greater than 60.
The load width act on pillar is distance apart two
tunnel. (W)
When distance of tunnel to ground surface (H) is
longer than tunnel width (D)
P= DW
: unit weight of ground
When distance of tunnel width (D) is longer than
tunnel to ground surface (H)
P= HW
1.
2.
3.
4.
5.
6.
Initial stage
Excavation completion of upper part of pilot tunnel.
Excavation completion of lower part of pilot tunnel
Load convergence
Excavation completion of precedence tunnel
Excavation completion of afterward tunnel
The load of act on pillar(KN)
account difference in behavior of joint that happen

in experiment and joint properties value that apply
in analysis. This is similar with suggested result by
Lee, kim. This result means that effect to support of
precedence tunnel by excavation of afterward tunnel. It
can become problem in stability secure of precedence
tunnel or support amount excess of afterward tunnel
that such conduct has precedence tunnel support and
afterward tunnel supports equally. Therefore hereafter
designs need to establish suitable support pattern plan
in conduct of 2-Arch tunnel.
35.0
Installation of Pillar
30.0
25.0
20.0
15.0
10.0
5.0
0.0
Figure 12. Change of the load acting on pillar at different

steps.
Table 6. The load acting on pillar (KN).
a
28.6
33.0
26.4
13.2
31.2
The maximum load acting on pillar of 2-Arch tunnel is 33.0 KN (Table 6). In Figure 12, the load of until
converge step that act on pillar of 2-Arch tunnel is
supported by rock bolt and shotcrete. The final load of
about 31.2 KN is the actual rock load to be supported
by the center pillar. In the current study, an empirical
relation for rock load based on the measurement is proposed for the rock when RMR is greater than 60. The
empirical equation proposed from the current research
is H = 0.15 W, where W is the centre-to-centre distance
between left and right tunnels. Although more study
is obviously required, this relation may be used for the
preliminary design of the centre pillar for the 2-arch
tunnel.
4
CONCLUSIONS
The 2-Arch tunnel is similar to two parallel tunnels

with very short centre-to-centre distance. However, it
674
is executing without enough studies for behavior of the

2-Arch tunnel. In this research, a study for behavior
of the 2-Arch tunnel is examined using large model
test machine. Then embody in-situ stress by applying
pressure to boundary of model ground.
1. The ground displacements mainly occurred within
0.25D, where D is the tunnel width. Horizontal displacements more than 40% of whole displacements
and vertical displacements more than 20% occurred
during excavation of pilot tunnel. Such result means
that stability of whole tunnel dominated by stability of pilot tunnel excavation in the 2-Arch tunnel.
This suggests that rock bolt length should be longer
than 0.25D to prevent the rock loosening.
2. Based on the limited information obtained from the
measurement, the rock load acting on the centre
pillar may be H = 0.15 W, where W is the centre-tocentre distance between left and right tunnels when
RMR is more than 60. However, obviously more
research is required to generalize the proposed
relation.
REFERENCES
Amadei, B. & Stephanson, O. 1997. Rock stress and its
measurement, Kluwer Academic Publishers.
Brady, B.H.G. & Brown, E.T. 1985. Rock mechanics for
underground mining, Kluwer Academic Publishers.
Goodman, R.E. 1988. Introduction to Rock Mechanics, John
Wiley and Sons.
Hoek, E. & Brown, E.T. 1980. Underground excavation in
Rock, Instn. of Mining & Metallurgy.
Hoek, E. 2000. Practical Rock Engineering, Course note.
Lee, S.D., Choi, S.I. & Gu, J.G. 1994. Design and construction of stability underground structure.
Lee & Kim. 2001. The consideration of improve 2-Arch
tunnel Design and construction method.
Lee, I.M. 2001. Principle of rock mechanics.
Matsuda, T. 1998. Ground behavior and Settlement control
of twin tunnels in soil ground. Tunnels and Metropolies:
pp11931198
Matsuda, T. 1998. Discussion behavior and settlement control
of twin tunnel in soil ground, Tunnels and Metropolises.
Priest, S.D. 1993. Discontinuity Analysis for Rock Engineering, Kluwer Academic Publishers.
Yun, J.S. 1991. Investigation and test of rock mass.
675
Behavior of tunnel due to adjacent ground excavation under the influence

of pre-loading on braced wall
S.D. Lee
Department of Environmental, Civil and Transportation Engineering, Ajou Univ, Suwon, Gyeongi-Do, Korea
I. Kim
Sambon Eng. Anyang, Gyeonggi-Do, Korea
ABSTRACT: Pre-loads could be imposed on the braced wall to prevent its horizontal displacements during
the ground excavation even though a tunnel exists adjacent to the braced wall. New pre-loading system for
large loads was developed and applied to the large scale model tests. Model tests were performed in the sandy
ground, which was homogeneously and isotropically constructed in the test pit and numerically analyzed by Finite
Element Method. It was found that the stability of existing tunnel was greatly enhanced when the horizontal
displacements of a braced wall was reduced by applying pre-load, which was larger than the design load.
INTRODUCTION
Horizontal displacement of a braced wall during the

ground excavation could soften the rear ground, which
could cause the existing structure in the rear ground to
be unstable (Lee, 1999). Pre-loading method is widely
used to reduce the lateral displacement of braced wall.
ORourke (1976) found that lateral displacement
of a braced wall could be reduced by pre-loading in
halfsize of the design force. But too large pre-loading
could jeopardize the strut. ORourke (1981) also proposed that the effective rigidity could be increased by
imposing pre-load.
To establish the optimum range of pre-loading
in clayey soil, Mana and Clough (1981) performed
numerical analyses and found that displacement could
be reduced by pre-loading the design force. But they
pointed out that too large pre-loading could cause
local deformation of the steel joint of the supporting
structure and could damage it.
Canadian Geotechnical Society (1997) found that
pre-loading which amounted to the design force
should be imposed to reduce the displacement of a
braced wall.
Most studies on the pre-loading have focused on the
ground behavior, the effective rigidity of strut, and the
displacement of a braced wall.
To keep the tunnel in the rear ground to be stable during the adjacent ground excavation, horizontal displacement of braced wall should be reduced.
Horizontal displacement, however, could not be sufficiently reduced by pre-loading the design force and it
is difficult to be convinced of the tunnel stability.
In this paper, behavior of the braced wall and the

tunnel in rear ground was studied, when pre-loading
was imposed on a braced wall during the ground excavation. For this purpose, large scale model tests were
conducted and numerically analyzed.
2
2.1
LARGE SCALE MODEL TESTS

Summary
Pre-loading method was verified through large scale

model tests. They were performed by imposing preload on a strut during the ground excavation by a new
pre-loading system, through which large preload could
impose.
Model tests were conducted in a large test
box (2.0 m wide, 4.0 m long, 6.0 m tall) under 2dimensional boundary conditions. Lateral wall of
test box was specially treated to minimize the friction. Existing tunnel was detached 0.5D from the
braced wall.
Until excavation was completed, wall displacement
was kept to be zero (18 stages) by pre-loading. Model
tests without pre-loading were also conducted.
2.2 Test ground
For model tests, homogeneous and isotropic model
ground was constructed. Sand was put into test pit
in 0.3 m depth at a time and compacted by a plate
vibrator. Relative density of test ground was checked
by extracting a sample using DIN 4021 Core Cutter
(Lee, 1998).
677
Figure 2. Grain size distribution curve.

Table 1.
Physical properties of test ground.
Parameter
Value
Max. Dry Unit Weight (dmax )

Min. Dry Unit Weight (dmin )
Dry Unit Weight of Test Ground (d )
Relative Density (Dr )
Water Content ()
Specific Gravity (Gs)
16.86 (kN/m3 )
13.82 (kN/m3 )
15.39 (kN/m3 )
56 (%)
6.8 (%)
2.63
Rigidity Ratio of Model Tunnel;
Figure 1. Model test box.
Physical characteristics of test ground for the model

tests were confirmed through grain size distribution
test, specific gravity test, field unit-weight test, and
water content test.
According to the Unified Soil Classification System (USCS), test ground was poorly graded Sand SP.
Mechanical properties of test ground were obtained
through a direct shear test. Internal friction angle ()
was 38 and cohesion (c) was 6.0 kPa. Modulus of
elasticity (E) was 28,000 kPa.
Test results are shown in Figure 2 and Table 1:
2.3
Model tunnel and model braced wall
Specifications of the model tunnel were determined

from the rigidity ratio by Duddeck and Erdmann
(1985). Rigidity ratio between the ground and the
tunnel lining for actual and model tunnel was as
follows.
Rigidity Ratio of Actual Tunnel;
Here, subscript m means model tunnel.

where Ek (Ekm ) =Youngs Modulus of Ground (kPa);
R (Rm ) = Radius of Tunnel (m); Eb (Ebm ) =Youngs
Modulus Lining (kPa); Ib (Ibm ) = Moment of Inertia of
Tunnel Lining; and
Thickness of model tunnel tm was estimated under
the assumption that the rigidity ratio of real and model
tunnel was equal (Soliman et al. 1993).
where t (tm ) = Thickness of tunnel lining (m)
Modulus of elasticity of the actual ground (Ek ) was

400,000500,000 kPa. But modulus of elasticity of
model ground (Ekm ) was 20,00030,000 kPa. For the
model tunnel lining with the diameter of 0.6 m, 6 mmthick steel was used, which was equivalent to 0.3 mthick concrete lining. For a braced wall, 16 mm-thick
steel was used, which was equivalent to an 0.8 m-thick
concrete slurry wall.
Hobbs (1966) estimated the time reduction rate
under the assumption that the acceleration of gravity
was constant. Reduction rate of length was established
first and then reduction rate for time, density, weight,
678
Table 2.
Table 3.
Reduction rate of model test.
Property
Dimensions
Reduction tate
(Tunnel lining
and braced wall)
Length
Time
Weight
Density
Stress
Acceleration
of Gravity
[L]
[T]
[M]
[ML3 ]
[ML1T2 ]
[LT2 ]
1/10
1/3.16
1/3.120
1/3.12
1/31.24
1.0
Properties of ground.
Parameter
Modulus of Elasticity (E)
Poissons Ratio ()
Unit Weight ()
Internal Friction Angle ()
Cohesion (C)
20,000 kPa
0.25
16.39 kN/m3
38
6.0 kPa
Figure 4. Finite element mesh and boundary conditions.
2.4 Measurement
Moment of braced wall (40 points), lateral displacement (8 points), rear ground surface settlement
(9 points), lining moment (32 points), and tunnel distortion (16 points) was automatically measured and
saved at every 30 minutes. Measuring points are shown
in Figure 3 and Table 3.
NUMERICAL ANALYSIS
3.1 Analysis program
Figure 3. Measuring points for model tests.
and stress was estimated. The law of similarity could

be based on weight, time, and acceleration. Steel with
unit weight 78 kN/m3 was used for model tunnel lining and the braced wall (Yang et al. 2007, Shim et al.
2007).
Tunnel was detached 0.5D from the braced wall.

Tests were numerically analyzed using FEM program
PLAXIS Ver. 8.2.
Triangular plane strain elements with fifteen nodal
points were used for the ground. Interface was adapted
on the boundary at the braced wall and the ground.
Finite element mesh and boundary conditions for
the numerical analysis are shown in Figure 4.
Vertical displacement at the lateral boundary and
horizontal displacement at the lower boundary were
possible.
679
Table 4.
lining.
Material properties for braced wall and tunnel
Parameter
Braced wall
Tunnel lining
Axial Rigidity (EA)

Flexural Rigidity (EI)
Thickness (d)
Poissons Ratio ()
3,293,000 kN/m
70.0 kNm2 /m
0.016 m
0.3
1,929,000 kN/m
5.788 kNm2 /m
0.006 m
0.3
Figure 5. Horizontal displacement of braced wall.
3.2
Figure 6. Distortion of tunnel lining.
Input data
Properties of ground and structure members used in

numerical analysis are shown in Tables 3 and 4:
4
4.1
RESULTS OF LARGE SCALE MODEL

TESTS AND NUMERICAL ANALYSES
Horizontal displacement of braced wall
Horizontal displacement of a braced wall in the large

scale model tests and numerical analyses is shown
in Figure 5. At the final excavation stage, maximum
lateral displacement (max) of braced wall decreased
from 2.290 mm to 0.03 mm by preloading, while the
result of numerical analysis decreased from 2.914 mm
to 0.193 mm (Fig. 5).
4.2 Tunnel distortion
Distortion of tunnel lining is shown in Figure 6.
Without preloading, tunnel distorted gradually due
to ground excavation. Tunnel was pressed vertically
and expanded horizontally. By pre-loading tunnel distortion greatly decreased. Maximum displacement of
the tunnel lining decreased from 2.22 mm to 0.21 mm
in the tests and from 2.670 mm to 0.337 mm in the
numerical analysis (Fig. 6) by pre-loading.
4.3
Member stress of tunnel lining
By pre-loading, maximum flexural moment decreased

from 0.15 kNm/m to 0.068 kNm/m and maximum
shear force decreased from 1.160 kN/m to 0.436 kN/m.

Axial force, however, increased from 17.436 to
22.837 kN/m in the numerical analysis (Fig. 6). Maximum flexural moment at the final excavation stage
decreased from 0.195 kNm/m to 0.021 kNm/m
by pre-loading while the shear increased from
0.331 kN/m to 0.178 kN/m. Maximum axial force
increased from 34.377 kN/m to 124.863 kN/m, in
the model tests.
In the numerical analyses, on the contrary, by
preloading, maximum flexural moment decreased
from 0.150 kNm/m to 0.068 kNm/m, maximum shear
decreased from 1.160 kN/m to 0.436 kN/m. Maximum axial force, however, slightly increased from
17.463 kN/m to 22.837 kN/m, which was in an
allowable range. Moment of the tunnel lining is shown
in Figure 7.
4.4 Lateral displacement of braced wall and
tunnel distortion
Without pre-loading, the largest displacement and tunnel distortion were observed in the 6th excavation
stage, in the tunnel level.
Significant horizontal displacement of braced wall
and tunnel distortion would not occur throughout
the entire process, when pre-loading was imposed. If
braced wall was not deformed laterally, tunnel shape
would hardly change (Fig. 8).
680
Figure 9. Rear ground surface settlement.
4.5 Ground surface settlement

By pre-loading maximum ground surface settlement
at final excavation stage decreased from 2.1 mm to
0.12 mm in the model tests and from 2.506 mm to
0.726 mm in the numerical analysis. Rear ground
surface settlement is shown in Figure 9.
Figure 7. Bending moment and stress of tunnel lining.
CONCLUSION
It was investigated how to ensure the stability of

the adjacent tunnel during the ground excavation
by reducing the horizontal displacement of a braced
wall by gradually imposing pre-loading. A new preloading system, through which large pre-loading
could impose was developed and applied to large
scale model tests. Performed tests were numerically
analyzed.
The results are as follows:
1 If lateral displacement of the braced wall was
greatly reduced by imposing pre-loading, which
was larger than design force, the stability of the
adjacent tunnel at the rear ground would greatly
increase.
2 Maximum flexural moment and maximum shear of
tunnel lining were decreased but the axial force was
increased by imposing pre-loading in both large
scale model tests and numerical analyses.
3 Largest horizontal displacement of a braced wall
occurred in the excavation stage, which was the
tunnel level, when pre-loading was not imposed.
4 Tunnel was pressed vertically and expanded horizontally during the ground excavation without
preloading. But by imposing pre-loading no significant distortion in its shape was observed throughout
the entire excavation process.
5 Most of the ground surface settlement was
decreased by imposing pre-loading.
Figure 8. Lateral displacement of braced wall.
681
REFERENCES
Canadian Geotechnical Society. 1997. Foundation Engineering Manual. 3rd. ed.
Duddeck, H. & Erdmann, J. 1985. On structural design models for tunnels in soft soil. Underground Space, Vol. 9,
Pergamon Press; 246259.
Hobbs, D.W. 1966. Scale model study of strata movement
around mine roadways. Apparatus, technique and some
preliminary results, Int. J. of Rock Mech. Min. Sci., 3.
Lee, S.D. 1998. Soil Mechanics (2nd. edition). Saeron Press;
353354.
Lee, S.D. 1999. Foundation Engineering. Saeron Press;
252254.
Lee, S.D. 2003. Understanding of the Latest Tunneling
Technology. Ajou University Geotechnical Engineering;
122.
Mana, A.I. & Clough, G.H. 1981. Prediction of movement for
braced cuts in clay. J. Geotech. Engineering. Div., ASCE,
vol. 107, No. 6; 759778.
ORourke, T.D., Cording, E.J. & Boscardin, M. 1976. The

Ground movements related to braced excavation and
their influence on adjacent buildings. U.S Department of
Transportation, Report no. DOT-TST 76, T-23.
O Rourke, T.D. 1981. Ground Movements Caused by Braced
Excavation. Journal of Geotechnical Engineering Division. ASCE. Vol. 107. NO. GT9; 11591178.
Shim, H.J. and others. 2007. Model Test and Test Blasting
to Design Adjacent Tunnel. KTA 2007 Annual Conference
April 2021 Seoul Korea; 267278.
Soliman, E., Duddeck, H. & Ahrens, H. 1993. two-and
three-dimensional analysis of closely spaced double-tube
tunnels. Tunnelling and Underground Space Technology,
Vol. 8, No. 1; 1318.
Yang, H.S. and others. 2007. A Study on Characteristics of
Materials for Scaled Model Test. Ministry of Construction & Transportation R&D Performance Forum Journal;
3336.
682
Two distinctive shear strain modes for pile-soil-tunnelling interaction

in a granular mass
Y.J. Lee
Soil-Structure Interaction Group, Steel Structure Research Laboratory, Research Institute of Industrial Science &
Technology (RIST), South Korea
C.S. Yoo
Department of Civil & Environmental Engineering, Sungkyunkwan University, South Korea
ABSTRACT: There are many high-rise buildings in urban areas, which are normally supported by piled foundations. Consequently, a major concern for geotechnical engineers is the construction of a tunnel adjacent to the
piled foundations, since ground behaviour between an existing loaded pile and tunnelling has not been understood well so far, particularly for granular soils rather than clay soils. In order to figure out such complicated
ground behaviour, the interactive shear strain patterns are generated by two-dimensional laboratory model tests
and finite element analyses (FEA). For the model testing, a multi-sized aluminium rod mixture considered as
a continuum granular mass and close range photogrammetric technique for obtaining displacement data within
the rods are introduced. Two distinctive shear strain modes, viz. connective and isolated modes, are presented
through the comparison of the model tests and FEA according to the pile tip locations.
INTRODUCTION
Tunnel excavation work in the soft ground results in

significant reduction in total stress in the vicinity of a
tunnel boundary. The reduction in stress causes ground
movements which affect adjacent building foundations
consisted of a row of loaded piles.
In order to identify the pile-soil-tunnelling interaction behaviour, small-scale physical model tests
and numerical methods were employed in this study.
Careful assessment of the pile-soil-tunnelling interaction problems is relatively new and only limited
information is currently available.
This study focuses on two distinctive shear strain
modes of the ground between the existing pile and a
new tunnel construction as shown in Figure 1. The laboratory model tests were conducted by strain control
rather than stress control, and then comparison with the
finite element analyses was carried out to identify the
shear strain modes. This research incorporates strain
controlled tests with idealised two-dimensional aluminium rods considered as a granular mass taken
to very high volume loss (up to about 20%) to highlight the full shear failure formation. Digital image
processing technique has allowed overall deformation
patterns of ground movements to be obtained from
the analysis of digital images. Detailed shear strain
patterns of the ground can be obtained which give a
clear insight into the pile-soil-tunnelling interaction

events.
The model pile was principally an end-bearing pile
where most of the pile load is concentrated on the pile
tip rather than pile shaft. The pile working load was
maintained constant during the test. This working load
was chosen by reference to a displacement-controlled
pile-load test (the pile working load, 3.6 kN, is 77% of
the ultimate pile load). In this study, in order to avoid
complexity of pile loading the influence of lateral
loading was not considered.
LABORATORY MODEL TEST
2.1 Test equipment

The displacement-controlled model tunnel consists of
6 segments, the two ends of a segment being carved
on tapered cones. Each segment moves inward as the
tapers are withdrawn, simulating the two-dimensional
volume loss during tunnelling operations. The tunnel
diameter is reduced by rotating the two knobs as shown
in Figure 2. The outer diameter of the tunnel is initially
100 mm. The reduction of the tunnel diameter gives
directly a 2D volume loss (VL ). This 2D volume loss
per revolution is determined from the calibration result
(Lee, 2004).
683
Figure 3. Multi-sized aluminium rod material.
Figure 4. Identification of pile tip location.
Figure 1. Schematic illustration of shear strain modes for

the pile-soil-tunnelling interaction.
The aluminium rod mixture consists of six different diameters (viz. 2 mm, 3 mm, 6 mm, 9 mm, 12 mm
and 20 mm), which have the same length of 75 mm.
It represents a well graded, idealised two-dimensional
granular material, as shown in Figure 3. The test chamber is rigid, a rectangular steel frame (width: 1058 mm,
height: 930 mm). The model pile (25 mm 75 mm in
cross section, embedded length, L: 370 mm) is made of
aluminium alloy. No significant effects of the boundary conditions were observed during the test, since the
smallest rods (2 mm and 3 mm diameters) were mainly
used in the interactive regions and the largest rods were
used in the vicinity of the steel frame boundaries.
2.2 Pile-soil-tunnelling interaction test
Figure 2. Diameter reduction system of model tunnel.
The pile tip identification system adopted was to label

the distances of the pile tip away from the tunnel centre
line (O) as A (1d0 : model tunnel diameter), B (1.5d0 ),
C (2d0 ), D (3d0 ) and the pile tip level as 0 (on the centre
line), +1 (on the crown level), +2 (at 1d0 above the
centre line level), 1 and 2 being similarly below the
centre line level. One final location (O + 2) was with
the pile directly above the centre line of the tunnel on
the +2 level (Figure 4).
684
Figure 6. Typical digital camera positions for capturing 10

images.
Figure 5. Set-up stage with test equipment and reflective

target points.
The pile was in place and loaded before any of the

pile-soil-tunnelling interaction occurred. The model
tests, therefore, simulated new tunnelling adjacent to
a row of existing loaded piles.
Each model test normally consisted of three stages
as follows: (1) set-up stage, material compacted as
before, with reflective nodes shown in Figure 5 (both
pile and tunnel installed in positions but with no pile
loading and tunnel at initial diameter, d0 ); (2) pile
loaded to working load by dead weight (Pw from the
pile-load test); (3) tunnelling stage (reduction of the
tunnel diameter to a maximum value of 12 revolutions,
i.e. VL = 18.65%).
The tests using aluminium rods in this study are
much quicker to carry out in terms of testing time
than any conventional test using a real soil such as
sand or clay. In addition, detailed observation of the
side section can easily be carried out, enabling details
of the failure mechanisms, which are associated with
strain fields rather than stress fields, to be examined.
CLOSE RANGE PHOTOGRAMMETRY
The close range photogrammetric technique for determining strains used in this study has recently been
applied to a number of geotechnical engineering problems. A Kodak DC 290 digital camera was used to
capture both the frame (or chamber) reference points
and the target points fixed to the centre part at the
ends of the smaller size rods within the multi-sized
rod matrix. The digital camera can provide acceptably
high resolution. Ten independent images were taken at
each epoch (or stage) using the Kodak DC 290 digital
camera as shown in Figure 6.
A pixel resolution of 1792 1200 was available.

Each retro-reflective target was identified and the position of its image within each of the ten photographs was
measured by the VMS (vision metrology system) program. The measured 2D x-y coordinates of the target
points from the VMS are arranged into a triangulation mesh by means of the EngVis program. Figure 7
shows an example of the imaging processing by the
VMS and the subsequent triangulation by the EngVis
respectively.
The system and its application were described by
Woodhouse et al. (1999), Woodhouse (2000) and
Kwok and Swajani (2001). From the measured displacements, strains were calculated based on the
assumption of linear strain in each triangular element
(Lee and Bassett, 2006). More detailed model test procedures matched with the image capturing stages are
shown in Figure 8.
NUMERICAL ANALYSIS
In order to carry out the finite element analyses (FEA),

the 3D pile-soil-tunnelling interaction situation was
idealised to two-dimensional plane-strain conditions
and to compare with the model pile-soil-tunnelling test
result, the same tunnel geometry, pile size, and location of the pile relative to the tunnel were adopted for
the FEA. The FE analyses were carried out using the
continuum finite element program CRISP (Britto and
Gunn, 1987; Woods and Rahim, 2001).
Ground behaviour was assumed to be governed by
an elastic-perfectly-plastic constitutive model based
on the Mohr-Coulomb failure criterion with a nonassociated flow rule. The critical state angle of friction
(cs ) and the angle of dilation () were determined to
be 23 and 15 for the model granular material, respectively (based on a best FEA fit of load-settlement
relationship for the model pile loading tests, Lee,
2004). It should be noted that later these parameters
were also obtained from shear box tests (area void
685
Figure 8. Relationship between model test steps and image

capturing stages.
Table 1.
Ground parameters used in the FEA.
Ground surface level, Y0 (m)

Youngs modulus at Y0 , E0 (kPa)
Gradient of Youngs modulus, m
E (kPa/m)
Poissons ratio,
Unit weight of soil, (kN/m3 )
Effective cohesion, c (kPa)
Critical state angle of friction, cs (degrees)
Angle of dilation, (degrees)
0.72
1600
10,000
0.35
24
0.1
23
15
Note: from bottom to top of mesh; varying with depth
Figure 7. Digital image analysis by VMS and EngVis

programs.
ratio, e = 0.34 0.38, cs and e values are found to

be similar to those given by Yamamoto and Kusuda,
2001). Parameter values from the shear box tests were
comparable to the best fit values. The effective cohesion (c ) was assumed to be 0.1 kPa. The variation of
Youngs modulus (E) was assumed to increase linearly
with depth. The ground parameters are summarised in
Table 1.
The tunnel support and pile was modelled as a twonode bar element and linear elastic material respectively. The Youngs modulus and Poissons ratio for
both the tunnel support and pile were assumed to be
15.5 GPa and 0.2, respectively. The unit weight of the
pile and the cross section area of the tunnel support
ring were 23 kN/m3 and 0.003 m2 , respectively. The
model parameters for the tunnel support and the pile
are summarised in Table 2.
Table 2.
Parameters for tunnel support and pile.
Parameters
Tunnel support
Pile
E (GPa)
Unit weight, (kN/m3 )

Cross section area, A (m2 )
15.5
0.2
0.003
15.5
0.2
23
FNR (Full Newton-Raphson) iterative solution

scheme was adopted together with a tolerance of 0.05
and a maximum iteration number of 100. K0 (0.66)
was applied as the initial in-situ stress conditions. Double convergence check based on both the displacement
and force norms and a total of 1055 increments were
used. It should be noted that the largest 2D volume loss
values (7.63% to 18.65%) were generated in order to
capture the interaction failure patterns between the pile
and the tunnel.
686
Figure 9. Comparison of maximum shear strain contours at

VL = 3.94%.
Figure 10. Comparison of maximum shear strain contours

at VL = 10.94%.
RESULTS
Figures 9 and 10 show the maximum shear strain

contours of the model tests at VL = 3.94% and
VL = 10.94% respectively. The location and intensity
of shear strain clearly identify the developing shear
failure formation between the pile base and the tunnel. It is noted that the case of O + 2 was omitted in
this comparison (this case is not a normal practice for
considering a proper tunnel position adjacent to a row
of loaded piles).
A clear neutral or dead block X an area with low
or no strain was observed clearly in case C + 1 at
10.94% of volume loss as shown in Figure 10.
The shear failure formation appeared to comprise two distinctive shear strain modes: (1) one that
includes a neutral soil block X, separating a formation running from the pile base to the tunnel invert area
and a second mechanism running from the pile shaft to
the tunnel crown area, and (2) the other one that is an
independent shear behaviour mode, i.e. no interactive
shear strain mode between the pile and the tunnel. In
summary, the two different shear strain modes shown
in Figure 1 can be identified according to pile tip
locations as shown in Figure 11.
687
of agreement in terms of shear strain data. Based on

the maximum shear strain data, it was observed that
the shear strain modes developed are strongly dependent on the pile tip location and the magnitude of the
volume loss. Through this study, it is recognized that
the boundary of two different shear strain modes may
be a useful guide for the tunnel planners who need
to make a decision on the proper positioning of tunnel
construction adjacent to a row of loaded piles in urban
areas.
REFERENCES
Figure 11. Boundary of two different shear strain modes

according to pile tip locations.
Similar shear strain mode behaviour was observed

at both small and large magnitudes of the volume loss.
While the model tests and FEA are twodimensional, it is expected that similar behaviour
would be found in real projects where the tunnelling
presents a three-dimensional problem. However, the
true boundary of two different shear strain modes in
the latter may be different from the former tests and
analyses.
6
Britto, A.M. & Gunn, M.J. 1987. Critical state soil mechanics via finite elements. Chichester, UK: Ellis Horwood
Limited.
Kwok, H.Y. & Swajani, C. 2001. Precise measurement technique. 3rd year project report, Department of Civil and
Environmental Engineering, University College London,
University of London.
Lee, Y.J. 2004. Tunnelling adjacent to a row of loaded piles.
PhD Thesis, University College London, University of
London.
Lee, Y.J. & Bassett, R.H. 2006. Application of a photogrammetric technique to a model tunnel. Tunnelling and
Underground Space Technology 21(1): 7996.
Woodhouse, N.G. 2000. Geometric models appropriate for
engineering analysis from vision metrology data. PhD
thesis, University College London, University of London.
Woodhouse, N.G., Robson, S. & Eyre, J.R. 1999. Vision
metrology and three dimensional visualisation in structural testing and monitoring. Photogrammetric Record
16(94): 625641.
Woods, R. & Rahim, A. 2001. SAGE-CRISP Technical Manual, Version 4. The CRISP Consortium Ltd.
http://www.mycrisp.com/demo/TECHMAN.pdf
Yamamoto, K. & Kusuda, K. 2001. Failure mechanisms and
bearing capacities of reinforced foundations. Geotextiles
and Geomembranes 19(3): 127162.
CONCLUSIONS
Comparison between the physical model tests and the

finite element analyses showed many successful points
688
Stability analysis of large slurry shield-driven tunnel in soft clay

Y. Li & Z.X. Zhang
Department of Geotechnical Engineering, School of Civil Engineering, Tongji University, Shanghai, P.R. China

INSA-Lyon, LGCIE, France
ABSTRACT: The possibility of partial failure of large slurry shield-driven tunnels is investigated by an upperbound approach in limit analysis and a three-dimensional numerical modeling for the Shanghai Yangtze River
Tunnel. The results of the upper-bound limit analysis failure mechanisms and the 3D numerical modeling have
shown that the partial blow-out of the upper part of the tunnel face occurs when the slurry pressure is too large
while the global collapse of the whole tunnel face occurs when the slurry pressure is too small. The failure
mechanisms and critical slurry pressures obtained from both approaches are presented and discussed.
INTRODUCTION
In recent years, the rapid growth in urban development has resulted in an increased demand for the
construction of tunnels for electric and communication cables, and transportation systems. For obvious
practical reasons such as accessibility, serviceability
and economy, these tunnels are constructed by shield
machines of large diameter and at shallow depths. The
Groene Hart Tunnel, constructed in 2005 in Netherlands, was carried out by a slurry-shield machine with
an outside diameter of 14.87 m. The M-30 Tunnel in
Madrid excavated by EPB shield machine, 15.2 m in
diameter, was until recently the biggest shield tunnel
completed in the world. In September 2006, two massive 15.43 m diameter slurry shield machines began
work on Shanghai Yangtze River Tunnel. With the
increase of the shield tunnel diameter, the excavated
volume is increased dramatically and the probability
of excavation in complicated stratum with different
types of soil layers increases greatly too. The stability
of the soil itself decreases at the same time. Thus, in
recent years, more and more attention was paid to the
face stability of large shield-driven tunnels.
The analysis of the face stability of shallow circular
tunnels driven by the pressurized slurry shield requires
the determination of the pressure to be applied by
the shield. This pressure must avoid both the collapse
(active failure) and the blow-out (passive failure) of
the soil mass near the tunnel face. A number of studies
have concerned tunnel face stability. Most results are
analytical and are based on limit equilibrium method

(Horn, 1961; Anagnostou & Kovri 1994; Broere,
2001) and limit analysis method (Davis et al. 1980;
Leca & Dormieux, 1990; Chambon & Cort, 1994;
Soubra 2000, 2002; Subrin & Wong, 2002). A rational and well-defined approach for the computation
of the supporting pressure is translational multiblock
failure mechanism of the upper-bound method, which
is presented by Soubra (2002) based on the 3D limit
analysis model of Leca and Dormieux (1990). This
mechanism allows the slip surface to develop more
freely in comparison with the available mechanisms
given by Leca and Dormieux, and thus, improves the
best upper-bound solutions given by these authors. The
multiblock mechanism is convenient for a constant
tunnel pressure which is an acceptable assumption for
small to medium tunnel diameters (10 m). However,
studies on face stability of very large slurry shielddriven tunnels, for which the hypothesis of a constant
slurry pressure is not applicable, are fairly few.
The non-constant supporting pressure of slurry
shield-driven tunnel is caused by the density of the
slurry. Since the density of the slurry should remain
within a certain range to obtain high-quality filter
cakes, and it is always smaller than the density of soil,
there will be a pressure difference between the slurry
pressure and the earth pressure at tunnel crown and
invert. This pressure difference increases with the tunnel diameter increase, shown in Figure 1, where, Su
and Sb are the slurry pressures at the level of the tunnel crown and invert; Pu and Pb are the corresponding
689
Figure 1. Pressure difference in different size tunnels.
Figure 3. Multiblock failure mechanisms.

Figure 2. Geological condition with C/D = 0.7.
earth pressures; D0 is the diameter of the small tunnel and D is the diameter of the larger one. This
pressure difference may induce a different failure
mechanism from the one corresponding to a constant
slurry pressure, especially in large tunnels.
In this paper, a simplified computation scheme considering the non-constant slurry pressure is adopted, in
which the multiblock failure mechanism suggested by
Soubra (2002) is employed to investigate the possibility of partial failure in large size slurry shield-driven
tunnel. Also, a more rigorous 3D numerical modeling is carried out to compare with the obtained results
of critical slurry pressures and the corresponding soil
mass at failure.
The tunnel is mainly excavated below the river bed

of Yangtze River, which is composed of muddy clay
and soft clay, with some local lenses of silty fine
sand. The tunnel will be driven at a depth up to 65 m.
A hydrostatic pressure up to 650 kPa is anticipated.
The shallowest ground section is under the river with
a cover-to-depth ratio of C/D = 0.7, where C and D is
the cover depth and diameter of the tunnel. Because
of the unfavorable geological condition and the large
dimension of excavating face, the face stability of the
tunnel is one of the key technical aspects in this project.
The most dangerous profile with C/D = 0.7 is chosen
for the study, as shown in Figure 2, where Hw is the
height of the water table.
3
2
CASE STUDY
PARTIAL FAILURE MECHANISM
3.1 Multiblock failure mechanism
The main part of Shanghai Yangtze River Tunnel is

7.5 km river-crossing tunnel connecting the Pudong
and Changxing Island in Shanghai, China. Excavation
of the tunnels is carried out by a pressurized slurry
shield machine with an outside diameter of 15.43 m,
which is the worlds largest until now.
The problem of the face stability analysis relevant to a

circular rigid tunnel of diameter D driven under a depth
of cover C could be idealized, as shown in Figure 3.
A surcharge s is applied on the ground surface, and t
is the uniform supporting pressure on the tunnel face.
The multiblock failure mechanism considered in this
690
paper is described in Soubra (2002). It is composed

of several truncated rigid cones with circular crosssections and with opening angles equal to 2, where
is the friction angle of the soil. The different blocks
of this mechanism move as rigid bodies. These rigid
cones translate with velocities of different directions,
which are collinear with the cones axes and make an
angle with the discontinuity surface. The velocity
of each cone is determined by the condition that the
relative velocity between the cones in contact has the
direction that makes an angle with the contact surface. The present mechanism is completely defined by
n angular parameters and i (i = 1. . .n-1), where n is
the number of rigid blocks. The geometrical construction of this mechanism is similar to that of Leca and
Dormieux (1990). The present mechanism is characterized by more freedom angles than that of Leca and
Dormieux (1990) and thus improves the solutions of
the critical supporting pressure.
The external forces contributing to the rate of external work consist of (i) the self-weight of the truncated
rigid cones; (ii) the surcharge loading s (in case of
outcrop of the upper rigid block) and (iii) the pressure
t at the face of the tunnel. The rate of energy dissipation occurs along the lateral surfaces and radial planes
of the failure mechanism. By equating the total rate
of external work to the total rate of internal energy
dissipation, the pressure t at the face of the tunnel is
obtained as follow:
where, Ns and N are surcharge and soil weight coefficient; c, are the cohesion and friction angle of the
soil and is the soil unit weight.
3.2 Critical slurry pressures
In order to study the possibility of partial failure,
two partial failure mechanisms are used as shown in
Figure 4, corresponding to the blow-out of the upper
part of tunnel and the collapse of the lower part of tunnel. The partial failure is assumed to occur in an area
with a vertical axis DL , where DL (0, D). In upper
part blow-out mechanism, as shown in Figure 4(a),
the top of the failure area passes through the tunnel
crown. In the lower part collapse mechanism, as shown
in Figure 4(b), the bottom of the failure area passes
through the tunnel invert. The upper part collapse and
lower part blow-out are not considered because they
are less dangerous.
The critical slurry pressure corresponding to partial upper part blow-out and lower part collapse are
computed in a simplified approach using the multiblock failure mechanism as follow: the tunnel pressure
obtained from the multiblock mechanism for different
Figure 4. Two kinds of partial failure mechanisms.
prescribed values of the tunnel diameters (corresponding to different values of DL in the present analysis)
are computed. The value of DL giving the minimal
(respectively maximal) blow-out (respectively collapse) pressure is considered as the critical pressure
causing a partial failure. A five-block model (i.e.
n = 5) is employed for this study (as shown in Figure 3)
since it was shown in Soubra (2002) that n greater
than 5 will not significantly improve the results. Water
above the ground surface is considered as a surcharge.
Because the excavation is executed quickly compared
to the soil consolidation, a typical undrained analysis
is employed. Table 1 summarizes the characteristics of
the soil. The unit weight of slurry is F = 12 kN/m3 .
In case 1, the cohesion is considered to be a constant.
In case 2, cohesion of the soil increases with depth
as cu = 0.95z + 0.4, where z is the depth of the soil
layer. The mean value from the level of tunnel invert
to the ground surface in case 2 is equal to the cu in
case 1. Case 2 is chosen to take into account the variation of the undrained cohesion between the tunnel
691
Table 1.
Soil parameters used in numerical modeling.
unit
weight cohesion
cu
0
(kN/m3 ) (kPa)
case1 18.2
case2 18.2
friction
angle
u
( )
24.5
0.01
0.95z+0.4 0.01
Youngs
modulus Poissons
E
ratio
(MPa)
3.21
3.21
0.495
0.495
crown and invert in extra large tunnel, especially in

normally consolidated clay.
According to the basic assumption of the multiblock
failure mechanism, the intersection of the failure block
and the tunnel face will be an ellipse with a long axis of
DL in vertical direction. The critical slurry pressure t0
obtained by multiblock model is a uniform pressure (as
shown in Figure 3). Considering the total force balance
on the failure face of diameter DL , an equivalent slurry
pressure is obtained. The slurry pressure t at tunnel
crown can be computed as follow: t = t0 s DL /2.
Actually, since the slurry pressure increase with depth,
the total force due to the slurry is acting on the tunnel
face with eccentricity e. When using uniform pressure t0 , a moment should be taken into consideration
at the same time. However, the moment is not considered here since the normalized eccentricity e/D is
very small in this study (about 2%4%).
The critical slurry pressure t is calculated for partial diameter DL varying in 1 m steps, which leads to
the maximum (or minimum) value of t , as shown in
Figure 5. In case 2, the cohesion for different DL is the
mean value from the top to the bottom of the failure
block. In upper part blow-out, cohesion decreases with
the decrease of DL , since the failure block becomes
more and more shallow. While in lower part collapse,
the failure block is always starting from the tunnel
invert to the ground surface, the mean value of cohesion for different DL is equal to the one in case 1.
Therefore, there is no difference between the value of
t in case 1 and 2 in collapse.
Figure 5 shows that in case 1, the maximum and
minimum critical slurry pressure is obtained when partial failure diameter DL = D = 15.43 m, which means
that the global failure will happen both in collapse and
blow-out cases.
In case 2, the minimum slurry pressure in blowout is obtained when partial failure diameter is 12 m,
which means that in blow-out failure, the partial failure
of upper part tunnel face will be more dangerous than
the global failure of the whole tunnel face. However,
there are no great differences among critical slurry
pressures when DL (10, 15.43). Therefore, in practice, global failure of the entire face and partial failure
with DL > 10 m will have the same probability. For collapse, the maximum slurry pressure is obtained when
Figure 5. Critical slurry pressure of partial failure mechanism.
DL = 15.43 m. With the decrease of the partial failure

diameter, the critical slurry pressure will decrease at
the same time: global failure of the whole tunnel face is
more dangerous than partial failure. The critical slurry
pressure at the tunnel crown level is 619.9 kPa and
265.9 kPa corresponding to blow-out (partially) and
collapse (globally), which is 712.5 kPa and 358.5 kPa
at the level of tunnel center.
NUMERICAL ANALYSIS WITH FLAC3D
4.1 FLAC3D numerical modeling

In order to investigate the behavior of the tunnel
face during failure, numerical analysis is carried out
with the commercially available finite-difference code
FLAC3D , which is an effective program for situations
where physical instability may occur.
In the model, due to symmetry, only one half is
included. The model is sufficiently large to allow
for any possible failure mechanism to develop and
to avoid any influence from the model boundaries
(as shown in Figure 6). The water table is 2D above
the tunnel crown. In order to focus the analysis on
692
Figure 7. Displacement contours when failure is observed

in case 2.
Figure 6. Studying profile and 3D numerical model.
the face failure in front of the shield machine, the

excavation process was simulated using a simplified
single-step excavation scheme, assuming that the tunnel is excavated 13 m (the length of the shield machine)
instantaneously. Such a simplified modeling scheme
had been successfully adopted in previous studies
(Gioda & Swoboda, 1999).
The different soil layers are assumed to be elastic
perfectly-plastic materials conforming to the MohrCoulomb failure criterion. An undrained analysis is
carried out by using the undrained parameters of the
soil as shown in Table 1.
The initial slurry pressure s0 is equal to the earth
pressure at rest at center of tunnel face (in this case,
s0 = 537.6 kPa). The slurry pressure is increasing
with the depth according to the density of the slurry.
Slurry pressure at the center of the tunnel face si will
be increased (or decreased) by multiplying s0 by a
pressure factor M in every construction phase i, until
blow-out (or collapse) occurs. The failure criterion is
defined as follows: the construction phase is considered as the beginning of failure, where for the first
time considerable value of unbalance force (not equal
to 0) and velocity (greater than e11 e12 m/step in
this case) is observed. Then the critical slurry pressure corresponding to collapse and blow-out will be
obtained by the following equation:
4.2
Numerical modeling results
For case 1, collapse occurs at the construction phase

when M = 0.65 and blow-out occurs at M = 1.30;
M = 0.70 and M = 1.30 for case 2 respectively. Maximum and minimum critical slurry pressure could be
obtained by equation (2). Displacement contours at
the failure of case 2 is plotted in Figure 7. A global
collapse failure of the whole tunnel face is observed.
This mechanism is well coincident with the centrifugal
experiment result of Chambon and Cort (1994). However failure mechanism of blow-out is a partial failure
Figure 8. Velocity contours of tunnel face when failure

occurs.
in the upper part of the tunnel face. Partial failure

mechanism is more obvious in case 2 than in case 1.
Two cases are compared in Figure 8. In collapse,
as shown in Figure 8 (a) and (b), the total tunnel face
will have considerable value of velocity; and the most
dangerous point with maximum velocity is near the
tunnel invert. The failure area on the tunnel face is the
whole face, which is circular in shape. In blow-out,
as we see in Figure 8 (c) and (d), the failure is constrained in upper 3/4D and 1/2D part of the tunnel face,
corresponding to the case 1 and case 2 respectively.
The failure mechanism in both cases has an elliptic
shape, with a long axis in the horizontal direction. The
most dangerous point with the maximum velocity is
the point near the tunnel crown. The failure mechanism
of case 2 could be obviously observed by the plot of
693
Figure 9. Velocity development of blow-out along the center

vertical axis of the tunnel face of case 2 (unit: m/step).
velocity evolution of the monitored points along the

center vertical axis of the tunnel face, as shown in
Figure 9.
COMPARISON OF FAILURE MECHANISMS
The failure mechanisms of five-block model and 3D

numerical modeling are compared in case 2 (as shown
in Figure 10). In five-block model, two sets of and
i are obtained by optimization of the coefficients
N and Ns . As shown in Figure 10, there are only
small differences between the two soil mass at failure
corresponding to the two sets of and i .
Basically speaking, the failure mechanisms of fiveblock model and 3D numerical modeling well agree
with each other. Both of them well predict the partial
failure on the upper part of the tunnel face in blow-out,
and global failure of whole tunnel face in collapse.
CONCLUSIONS
1. Both the results of partial failure mechanism based

on multiblock model and 3D numerical analysis
show that the partial failure mechanism will happen in blow-out but global failure with entire tunnel
face will dominate the collapse failure, especially
in case where cohesion is changing with depth.
Figure 10. Comparison of failure mechanisms of Case 2

(velocity contour for FLAC3D analysis).
Table 2. Comparison of critical slurry pressure at tunnel
center level in case 2.
Critical slurry
pressure /kPa
Pressure factor
M
Collapse Blow-out Collapse Blow-out

5-Block Model 358.5
3D Numerical 376.3
712.5
698.9
0.67
0.70
1.33
1.30
2. The critical slurry pressures of both multiblock partial failure mechanism and 3D numerical analysis
are as shown in Table 2. The results well agree with
each other. It should be noted that the multiblock
model is the upper-bound solution; it is possible
that a smaller value of the critical slurry pressure
for blow-out and greater one for collapse could be
found.
694
3. The results of 3D numerical analysis show that,

in blow-out, the failure area on the tunnel face is
described by an elliptic shape, with a long axis
in horizontal direction, and a short axis in vertical
direction. However, this shape of failure mechanism
is different from the assumption of the multiblock
model of upper-bound theorem, which is an elliptic
shape with long axis in vertical direction.
4. In multiblock model analysis of case 2, a mean
value of cohesion cu from the top to the bottom
of the failure block is employed. However, according to the shape of the failure block, the changing
cohesion will influence the energy dissipation along
the lateral surface of failure block. A more precise
calculation of energy dissipation along the failure
surface with cu changing with depth is necessary
to take into account. Also, the rotation of failure
block caused by the eccentricity of the slurry pressure should be considered in multiblock model to
obtain better results.
ACKNOWLEDGEMENTS
The research were conducted with funding provided by
the National High Technology Research and Development Program (863 Program) of China and Shanghai
Leading Academic Discipline Project, Project Number: B308. The first author is grateful to EGIDE to
provide the French funding scholarship of her doctoral
stay in INSA Lyon. Particular thanks are due to Prof.
Soubra. A. H. for important discussion and the use of
multiblock software.
REFERENCES
Anagnostou, G. & Kovri, K. 1994. The Face Stability
of Slurry-Shield-Driven Tunnels. Tunnelling and Underground Space Technology. Vol.9, No. 2:165174.
Broere, W. 2001. Tunnel Face Stability & New CPT Applications. PhD thesis, Delft University of Technology. Delft
University Press, the Netherlands.
Chambon, P. & Cort, J.F. 1994. Shallow tunnels in cohesionless soil: Stability of tunnel face. ASCE Journal of
Geotechnical Engineering. 120: 11481165.
Davis, E.H., Gunn, M.J., Mair, R.J. & Seneviratne, H.N.
1980. The Stability of Shallow Tunnels and Underground
Openings in Cohesive Material. Geotechnique. 30 (4):
397416.
Gioda, G. & Swoboda, G. 1999. Developments and Applications of the Numerical Analysis of Tunnels in Continuous Media. International Journal for Numerical and
Analytical Methods in Geomechanics. 23: 13931405.
Horn, M. 1961. Horizontal earth pressure on perpendicular tunnel face. Hungarian National Conference of the
Foundation Engineer Industry, Budapest. (In Hungarian)
frictional material. Gotechnique 40, 4: 581606.
Soubra, A.H. 2000. Three-dimensional face stability analysis of shallow circular tunnels. International Conference
on Geotechnical and Geological Engineering, 1924
November 2000. Melbourne, Australia.
Soubra, A.H. 2002. Kinematical approach to the face stability analysis of shallow circular tunnels. 8th International Symposium on Plasticity, 443445. Canada, British
Columbia.
Subrin, D. & Wong, H. 2002. Tunnel face stability in frictional
material: a new 3D failure mechanism. C. R. Mecanique,
330: 513519. (In French)
695
Effects of soil stratification on the tunneling-induced ground movements

F.Y. Liang, G.S. Yao & J.P. Li
ABSTRACT: The available analytical predictions for tunneling-induced ground movements are usually based
on the assumptions that the ground is homogeneous and elastic. Actually, effects of soil stratification should be
taken into account. In order to clarify this problem, a FLAC3D computational model was conducted to investigate
the effect of vertical stratification in soil on the tunneling-induced ground movements. The applicability of the
presented model is verified with other available published results as well as the field case histories. Surface
settlements and lateral displacements induced by tunneling with layered soil model were compared with those
based on homogeneous soil. The results show that the stratification of soil should not be neglected in the analysis
of the tunneling-induced ground movements, especially when the soil is assumed to be an elastic medium.
1
INTRODUCTION
Tunneling process will inevitably result in ground

movements, and as a result tunnel constructions in
urban area may cause serious damage to the adjacent
preexisting buildings or public facilities. Therefore
one of the important issues of tunneling in urban
areas is the estimation of potential ground movements. Methods for estimating the ground deformations due to tunneling may be classified into three
categories: empirical methods, numerical methods,
and analytical methods.
In the engineering practices, ground deformations
are often described as a normal distribution curve with
empirical formulas based on field observations (e.g.,
Peck 1969; New & OReilly 1991). These methods
are limited in their applicability to cope with different ground conditions and construction techniques,
and the information they can provide is limited in the
horizontal movements and subsurface settlements.
Some attempts have been made to develop simple
closed-form analytical solutions for tunnelinginduced ground movements in clays with the assumption of a uniform radial or oval-shaped ground
deformation pattern around the tunnel section (e.g.,
Sagaseta 1987; Verruijt & Booker 1996; Loganathan &
Poulos 1998; Park 2004). However, the effect of soil
stratification was not considered in these analytical
methods.
Based on the three-dimensional (3-D) nature of the
stress changes and deformations, numerical methods
such as finite element method and boundary element method could be used to analyze the effect of
soil stratification on the tunneling-induced ground
movements (Mair et al. 1996) and 3-D finite element

simulation models for shield-driven tunnel excavation,
taking into account relevant components of the construction process, have been developed (e.g., Lee &
Rowe 1991; Rowe & Lee 1992).
In order to analyze the effects of soil stratification on tunneling-induced ground movements, a 3-D
computational model was used to simulate the tunnel
excavation in soft soil in this paper. Firstly, the factors
of the computational model taking into account the soil
stratification were described. Secondly, the simulation
of the Thunder Bay Tunnel in layered soft soil using
FLAC3D with Mohr-Coulomb material model was
carried out. The predicted ground movements were
investigated in detail. In order to check the applicability of the proposed computational model, the results
were compared with those from Lee & Rowe (1991)
and the field data observed by Belshaw & Palmer
(1978). In the end, another simulation for the tunnel
using the elastic model was carried out. The results of
the ground movements were compared with the analytical results of Loganathan & Poulos (1998) and the
field data observed by Belshaw & Palmer (1978) and
Lee et al. (1992). The effects of soil stratification were
discussed and some suggestions for future research
were also put forward.
2
2.1
METHOD OF ANALYSIS
Simulation of the ground loss
The settlements caused by tunneling are often characterized by the term ground loss. The main cause
to the ground loss is that there is the gap between the
697
Figure 1. Definition of GAP (Lee et al. 1992).
construction section and the design section. Then the

actual ground movement at the tunnel opening section
is oval shaped.These movements can be approximately
incorporated to the volume of ground loss through the
heading and over the shield (as shown in Figure 1).
Thus the gap parameter (GAP) can be considered as
the maximum settlement at the tunnel crown, and it
may be expressed as Lee et al. (1992).
where GP = physical gap (GP = 2 + ) that represents the geometric clearance between the outer skin
of the shield and lining; = thickness of the tailpiece; = clearance required for erection of the lining;
u3D
= equivalent 3-D elasto-plastic deformation at the
tunnel face; and = value that takes into account the
quality of workmanship.
Applying tail void grout will lead to a much smaller
value of GAP than indicated here, but this value is a
possible maximum.
2.2 Principal assumptions
Figure 2. 3-D model geometry used in analysis.
3. Uniform pressure is applied to the ground surface

as the external loads. The ground stress field is
generated by the gravity force.
4. The medium is simplified as the continuum, ideal
elasto-plastic material, and the Mohr-Coulomb
material model is adopted.
5. The oval-shaped ground deformation pattern is
imposed as the boundary condition at the tunnel
opening.
EFFECTS OF SOIL STRATIFICATION

USING ELASTO-PLASTIC MODEL
3.1 Calculation model and boundary conditions
The medium around the tunnel almost always exists

in the form of stratification features, and the tunneling analysis is a matter of three-dimensions rather
than two-dimension. FLAC3D is used to simulate the
tunneling-induced ground movements in this paper.
FLAC3D is a finite difference three-dimensional software package based on the Lagrangian difference
method. FLAC3D can offer an ideal analysis tool for
solution of three-dimensional problems in geotechnical engineering. The assumptions of the analytical
method are presented in brief as follows.
1. A plane contact on the interface of different soil layers is assumed. The transverse movements between
the soil layers are neglected, and then the friction
force of the soil layers can be ignored.
2. The inner material is homogeneous in the same
soil layer. The influence of the ground water is
neglected.
A 3.3-km-long, 2.47-m-diameter section of the sanitary trunk sewer tunnel in the city of Thunder Bay,
Ontario, Canada, was constructed in 1976. The tunnel
was constructed through soft clay using a tunnelboring machine together with a segmented precast
concrete tunnel lining. The details of tunnel construction, soil condition, and the tunnel dimensions were
published elsewhere by Belshaw & Palmer (1978) and
Lee & Rowe (1991).
Because of the symmetry of the structure and the
influence range of the tunnel excavation, only half the
tunnel is considered. The depth from the ground surface to the tunnel axis is 10.7 m. The model geometry
is 30 m long (y direction), 25 m deep (z direction) and
20m wide (x direction). Figure 2 illustrates the 3-D
mesh used in this analysis. There are 21240 elements
and 23464 nodes. The lateral and bottom surface of
the model are displacement boundaries. The model is
698
fixed in the horizontal direction at lateral surface, and

the vertical displacement is limited at the bottom surface. The top surface of the model is the ground surface
and is free in all directions.
For the Thunder Bay tunnel, the estimation of the
gap parameter proposed by Lee et al. (1992) has been
used and the physical gap GP is 90 mm.The calculation
has two stages. Firstly, the radial oval displacement as
a boundary condition is applied on the tunnel crown.
The 3D movements ahead of the face into the zone to
be excavated can then be determined, and then u3D

can
be evaluated and used to calculate the total gap GAP.
Secondly, the model based on the tunnel diameter
2R + GAP, is established. The radial oval displacement as a boundary condition is applied on the tunnel
crown. Hence the ground settlements and the lateral
displacement around the tunnel can be simulated.
In this study, the structural behavior of the tunnel
shield and lining are neglected. In order to ensure the
accuracy of the simulation, three cases are assumed.
1. The tunnel is assumed to be excavated under perfect alignment, and the tunnel machine is pressed
hard against the face so that 3-D movement will be
minimized.
2. Full release of axial stress at the tunnel face is
assumed, which would result in the full development of 3-D ground loss ahead and over the tunnel
shield.
3. The excavation of the tunnel is the instantaneous
excavation, and the interrelationship between the
supporting and lining of the tunnel is neglected.
Results of the ground movements (surface and
sub-surface settlements, lateral displacement) from
the numerical simulation are defined as the final
ground movements after the tunnel excavation.
3.2
Comparison with observed results and other

numerical methods
In this section, the soil is assumed to have an elasticperfectly plastic constitutive relationship and a MohrCoulomb material model. The silt and silty sand strata
for the tunnel are divided into four sublayers according to the data in Lee & Rowe (1991). The layers were
assigned soil parameters on the basis of soil densities, laboratory testing, and empirical correlations with
SPT data. The parameters adopted for these materials
in the present analysis are based on the research of
Lee & Rowe (1991) and summarized in Table 1. The
material parameters for the uniform soil model are the
average values of the four layers respectively.
Results from the presented analysis were compared
with those from Lee & Rowe (1991) as well as field
measurements recorded by Belshaw & Palmar (1978).
The analysis model used by Lee & Rowe (1991) was
also based on elasto-plastic constitutive relationship.
Figure 3. Surface settlement troughs at various distances

from tunnel face.
3.2.1 Surface settlement along transverse section

The development of transverse surface settlement
troughs (i.e., settlement profile along the ground surface perpendicular to the direction of tunneling) for
various calculation methods at 2.2 m behind the tunnel
face are shown in Figure 3.
The shape of surface settlement troughs calculated
with layered soil model is consistently slightly wider
than the observed troughs, and its values are approximate to the results from Lee & Rowe (1991). From
the curves, the settlement values based on the uniform
soil model slightly underestimate the values based on
the layered soil. Figure 3 shows that consideration in
lamination of the soil has only a small effect on the
calculated surface settlement trough.
3.2.2 Variation of lateral displacement
The lateral displacement values when the tunnel face
was 15 m away from the monitoring points are shown
in Figure 4. The monitoring data are 2.2 m away from
the vertical centerline of the tunnel. The present results
based on the layered soil are very similar to those
obtained by Lee & Rowe (1991) using the 3-D finite
element analysis. The results from the two methods are
in good agreement with the field data. From the curves
in Figure 4, the horizontal movement values based
on the uniform soil model greatly underestimate the
field data above the tunnel horizontal centerline, and
the latter are almost 1.5 8 times larger than the former. Under the tunnel horizontal centerline, the lateral
displacements based on uniform soil model decrease
rapidly and are close to those based on layered soil
and the field data. As well known, the lateral displacement above the tunnel centerline has an important
influence on the movement of structures located close
to the tunnel. Figure 4 shows that consideration of
soil lamination has a significant effect on the lateral
displacement.
699
3.2.3 Variation of subsurface settlement with depth

above the tunnel axis
The observed and calculated subsurface settlement
distributions with depth at 15 m from the tunnel face
are shown in Figure 5. Results based on the layered
soil are very close to the results obtained by Lee &
Rowe (1991), and they are also in good agreement
with the field data. The settlement values based on
the uniform soil model underestimate the field data in
the whole measurement field, and the latter are almost
1.06 1.30 times larger than the former.
From the above analyses, results from this model

taking into account the stratification of the soil are in
good agreement with those obtained by Lee & Rowe
(1991) as well as the field data, and more accurate than
those based on the uniform soil model. Therefore the
FLAC3D model established in this paper proves to be
reasonable and reliable for the simulation of tunnelinginduced ground movement.
4
EFFECTS OF SOIL STRATIFICATION

USING ELASTIC MODEL
Loganathan & Poulos (1998) redefined the ground

loss parameter with respect to the gap parameter
and incorporated this into the closed form solution
derived by Verruijt & Booker (1996). The proposed
approach tended to overpredict the ground loss parameter for tunnels in soft clay, and the predicted surface
settlement troughs were wider than the field observations. Nevertheless, in general, the observed and
calculated settlement and horizontal movement are in
good agreement for tunnels in uniform clay. However,
the applicability of the solution based on uniform soil
model should be discussed when it was applied to
layered soils.
In this section, the soil is assumed to have a linear
elastic constitutive relationship. The ground around
the tunnel is also divided into four sublayers. The
parameters adopted for these materials in the present
analysis are summarized in Table 1. The material
parameters of uniform soil model are the average
values of the four layers respectively.
Results from the present analysis are compared with
those from Loganathan & Poulos (1998) as well as
field measurements recorded by Belshaw & Palmar
(1978) and Lee et al. (1992). The effects of the stratification of the soil on the ground movements are
analyzed in the following.
4.1 Surface settlement along transverse section
Figure 5. Subsurface settlements with depth above tunnel

axis.
Table 1.
Settlement results from this calculation, Loganathan &

Poulos (1998) and the field data (Lee et al. 1992) are
shown in Figure 6. Apparently, the values predicted
by Loganathan & Poulos (1998) overestimate the field
Mechanical parameters of materials.
No.
Description of soil layers
Thickness (m)
(kNm3 )
Eu (MPa)
Eb (MPa)
c (kPa)
(o )
K0
1
2
3
4
Peat
Loose silty sand
Silty clay
Varved clay
1.0
7.0
5.2
11.8
14.00
19.77
17.45
18.70
12.3
15.4
9.2
32.0
213.0
119.3
110.5
240.0
39
40
29
59
30
32
26
34
0.5
0.52
0.88
0.8
Notes: unit weight of soil; Eu shear modulus; Eb bulk modulus; ccohesive strength;
internal friction angle; K0 coefficient of static earth pressure.
700
Figure 6. Surface settlement troughs at various distances

from tunnel face.
data. The values based on the layered soil underestimate the values predicted by Loganathan & Poulos
(1998) and are close to the field data despite the fact
that they are slightly greater than the field data. From
the curves, the settlement values based on the uniform
soil model are greater than the values based on the
layered soil in most areas and the trough of the former is obviously wider than that of the latter. But the
maximum settlements obtained by this model underestimate the value obtained by Loganathan & Poulos
(1998) and the field data at the tunnel crown.
4.2 Variation of lateral displacement
The values of the lateral displacements monitoring
points are shown in Figure 7. The values obtained by
Loganathan & Poulos (1998) overestimate the field
data (Belshaw & Palmer 1978) within the depth range
from 5 m to 15 m. The values based on the layered soil
are in good agreement with the field data. From the
curves, when adopted the uniform soil model, the values are less than the field data at the range of depth
from 0 to 7 m, and close to the field data at the range
of depth from 7 m to 11 m and greater than the field
data at the range of depth from 11 m to 25 m. Figure 6
shows that the prediction of the ground lateral displacements based on the layered soil model is better
than the uniform model.
4.3 Variation of subsurface settlement with depth
above the tunnel axis
The results of the subsurface settlement above the
tunnel centerline are shown in Figure 8. The values
obtained by Loganathan & Poulos (1998) underestimate the field data (Lee et al. 1992) greatly in spite
that the trends of both profiles are similar. The present
results based on layered soil show larger values for
the surface settlement than those from Loganathan &
Poulos (1998) and close to the field data except that
Figure 8. Subsurface settlements with depth above the

tunnel axis.
there is an error of 7 11 per cent at the range of depth

from 0 to 2 m. At other depths the error is less than 5
per cent. The subsurface settlement values based on the
uniform soil model obviously underestimate the field
data and there is an error of 20 30 per cent between
them.
CONCLUSIONS
A FLAC3D computational model for simulating the

displacement caused by excavation of the Thunder Bay
tunnel has been performed, and the effects of the stratification of the soil on the tunneling-induced ground
movements are analyzed sufficiently.
Results from the model based on layered soils are in
good agreement with those obtained by Lee & Rowe
(1991) as well as the field data. Comparatively, results
disregarding the stratification of the soil will have distinct errors. Consequently, in engineering practice, the
stratification of soils should not be neglected in the
analysis of the tunneling-induced ground movements.
701
Furthermore, the results based on the elasto-plastic

model can simulate the deformation curves better than
those based on the elastic model.
Based on the elasto-plastic model, the analysis
results showed that the soil stratification has very few
effects on the ground surface settlement, but it has
greater effects on the ground lateral displacement and
sub-surface settlement.
While based on the elastic model, the analysis
results showed that the soil stratification has some distinct effects in the ground surface settlement, lateral
displacement and sub-surface settlement respectively.
In order to obtain more reasonable analytical solution to the tunneling-induced ground movements, it
is necessary to incorporate the stratification of soils
into the analytical solution proposed by Loganathan &
Poulos (1998).
ACKNOWLEDGEMENTS
This work reported here is partly supported by National
Natural Science Foundation of China (Grant No.:
50708078) and Program for Young Excellent Talents
in Tongji University (Grant No.: 2006KJ043). The
authors wish to express their gratitude for the financial
assistances.
Lee, K.M. & Rowe, R.K. 1991. An analysis of threedimensional ground movements: the Thunder Bay tunnel.
Canadian Geotechnical Journal 28: 2541.
Lee, K.M., Rowe, R.K. & Lo, K.Y. 1992. Subsidence owing
to tunnelling. I. Estimating the gap parameter. Canadian
Geotechnical Journal 29: 929940.
Loganathan, N. & Poulos, H.G. 1998. Analytical prediction for tunnelling-induced ground movements in clays.
J. Geotech. Geoenviron. Eng. 124(9): 846856.
Mair, R.J. 1996. Prediction of ground movements and assessment of risk of building damage due to bored tunneling. Proceedings of geotechnical aspect of underground
construction in soft ground, Rotterdam, Balkema Press:
713718.
New, B.M. & OReilly, M.P. 1991. Tunnelling induced ground
movements, predicting their magnitude and effects. Proc.,
4th Conf. on Ground Movements and Structures, Cardiff,
Wales, Pentech Press: 671697.
Park, K.H. 2004. Elastic solution for tunnelling-induced
ground movements in clays. Int. J. Geomech., 4(4):
310318.
Peck, R.B. 1969. Deep excavations and tunnelling in soft
ground. Proc., 7th Int. Conf. on Soil Mech. and Found.
Engrg.: 225290.
Rowe, R.K. & Lee, K.M. 1992. Subsidence owing to tunnelling. II. Evaluation of a prediction technique. Canadian
Sagaseta, C. 1987. Analysis of undrained soil deformation
due to ground loss. Geotechnique 37: 753756.
Verruijt, A. & Booker, J.R. 1996. Surface Settlements
Due to Deformation of a tunnel in an elastic half
plane.Geotechnique 46(4): 753756.
REFERENCES
Belshaw, D.J. & Palmer, J.H.L. 1978. Results of a program of
instrumentation involving a precast segmented concretelined tunnel in clay. Canadian Geotechnical Journal 15:
573583.
702
Centrifuge modelling to investigate soil-structure interaction mechanisms

resulting from tunnel construction beneath buried pipelines
A.M. Marshall & R.J. Mair
Department of Engineering, University of Cambridge, Cambridge, UK
ABSTRACT: New underground construction is undertaken increasingly close to existing buried structures.
The resulting effects on these structures must be properly evaluated. This paper examines the case of tunnel
construction built transversely to existing buried continuous pipelines using data obtained from tests within the
University of Cambridge Geotechnical Beam Centrifuge. This research aims to visually validate hypothesized
soil-structure interaction mechanisms that account for pipeline behaviour. This is accomplished by placing the
tunnel-soil-pipeline system directly against a Perspex wall within the centrifuge package so that digital images
can be taken of the soil and buried structures. Particle Image Velocimetry (PIV) is used to measure displacements
and provide a complete description of the soil-structure interactions.
INTRODUCTION
Beneath the surface of any major city is an intricate

and increasingly congested series of tunnels, pipelines,
and buried structures. New tunnels are constructed
for various purposes and form an important part of
urban infrastructure. Tunnel construction has varying
effects on surrounding ground depending on soil type
and construction process (Mair and Taylor, 1997). The
design of new tunnels must account for likely effects
on nearby buried structures.
Analytical methods for this problem typically fall
within three categories: [1] finite element (FE) or
finite difference (FD) methods, [2] Winkler-type models, and [3] continuum solutions. Winkler models and
continuum solutions generally provide quick solutions
(see for example Attewell et al., 1986; Klar et al.,
2005; Vorster et al., 2005; Klar et al., 2007) however
they all have similar simplifying assumptions that limit
their applicability. Klar & Marshall (2007) validated
some of the simplified assumptions of the continuum
method by comparison with more rigorous shell structure solutions. FE and FD methods are best suited for
the problem because they can incorporate the complex
tunnel-soil-pipeline interactions. For these models to
be applicable, they must be validated against real soil
and soil-structure behaviour.
Vorster et al. (2005) proposed a series of global
and local soil deformation mechanisms to account
for observed pipeline behaviour above a tunnel during centrifuge testing. The proposed mechanisms are
based on deformation and stress measurements made
within the soil mass.
Figure 1. Centrifuge package.
The current research aims to elucidate the soilstructure interaction mechanisms that account for the
behaviour of existing tunnels and buried pipelines
located above new tunnel construction by providing
visual evidence of soil and structure deformations.
This is accomplished by performing similar tests to
those described in Vorster et al. (2005) but with a new
centrifuge package that places the tunnel-soil-pipeline
system adjacent to a Perspex wall such that digital
images of the system can be analyzed using Particle
Image Velocimetry (PIV).
2
CENTRIFUGE PACKAGE
A centrifuge package was developed that allowed

visual observations to be made of the soil, model
703
Figure 2. Schematic of centrifuge package.
tunnel, and buried pipeline. In doing so, digital images

could be taken during the test and PIV (White et al.,
2003) could be used to accurately measure soil and
structure displacements.
Figures 1 and 2 present the developed centrifuge
package. The relevant components are labelled as
follows:
A: The model tunnel: composed of a stiff inner
cylinder, fixed within the box walls, and sealed within
a flexible rubber membrane. Fluid is extracted from
within the sealed rubber membrane to replicate tunnelling volume loss;
B: Three 8 megapixel digital cameras. These are
used to photograph the soil and buried structures for
PIV analysis;
C: Five lasers used to measure surface vertical
displacements within the middle third of the box
where boundary effects are negligible (see Figure 2
for locations);
D: Five linear variable differential transformers
(LVDTs) used to measure subsurface vertical displacements within the middle of the box and at a depth of
70 mm, corresponding to the axis of the model pipeline
(see Figure 2 for locations);
E: The model pipeline: consists of an aluminium
half-cylinder with a length of 700 mm, an outer radius
of 9.5 mm, and a wall thickness of 1.6 mm.
All tests were performed at 75 g using dry Leighton
Buzzard fraction E sand at a relative density of
approximately 90%.
BASELINE TESTS NO PIPELINE
Two tests were completed in which no model pipeline

was included. These tests served two purposes: first, to
investigate the effects of placing a layer of glass at the
Perspex interface (thought to reduce boundary effects)
and second, to obtain baseline displacement data for
the simulation of tunnelling in sand.
Figure 3. Vertical soil displacement data for a) surface measurements with glass interface, b) sub-surface measurements
with glass interface, c) surface measurements with Perspex
interface, and d) sub-surface measurements with Perspex
interface.
Volume loss (VL) is used as a reference throughout the following text. Two methods of determining
volume loss are referred to and are defined as follows:
1 Tunnel volume loss: the change in tunnel volume (or volume of fluid extracted from the
model tunnel) divided by the original total tunnel volume, expressed as a percent (VLtunnel =
Vtunnel /Vtunnel 100)
2 Soil volume loss: calculated by integrating the soil
settlement profile at a given depth and dividing by
the original tunnel volume, expressed as a percent
(VLsoil = Vtrough /Vtunnel 100).
3.1 Assessment of boundary friction using glass
and Perspex
Boundary effects are an issue when obtaining PIV data
from centrifuge tests. An assessment of the boundary
effects on the obtained PIV results was carried out by
comparing the PIV displacement data to that obtained
from the lasers and LDVTs which were placed within
the middle third of the thickness of the box (where
boundary effects are negligible). Glass is often used as
an interface between the soil and the Perspex wall when
performing PIV tests because it is harder than Perspex
and therefore prevents sand grains from scraping into
the Perspex at high stress levels (a phenomenon which
causes increased boundary friction).
Two tests were carried out: one with glass and one
with Perspex alone. Figure 3 presents a comparison
between the PIV centreline data (directly above the
tunnel) and the surface (laser) and sub-surface (LVDT)
data, for both the glass and a Perspex interface tests.
Figure 3 shows very good correlation between the
704
PIV data and that obtained from measurements made

within the middle of the box for both glass and Perspex interfaces, indicating very little friction at the
interface.
The use of glass caused optical problems (coloured
zones caused by refraction of light) which affected the
quality of the PIV analysis. It was therefore decided
that glass would not be used since the PIV displacement data was effectively the same for glass and
Perspex.
The reason for the departure of the PIV data from the
general trend in Figure 3b) at a tunnel volume loss of
about 3.7% is not known. It may be a result of soil loss
between the glass and the Perspex around the model
tunnel. This potential problem was addressed in subsequent tests by sealing circumference of the tunnel
ends with grease.
3.2 Tunnelling in sand
The baseline tests provided useful data regarding the
effects of tunnelling in sand. Figure 4 presents the variation of volume loss at various depths within the soil
for a) the baseline test with a glass interface, and b)
the baseline test with a Perspex interface. The 1:1 line
represents the volume loss of the model tunnel. When
the slope of a line is greater than 1, the soil is experiencing a volume loss that is larger than that provided
by the volume loss in the tunnel.
Figure 4 illustrates that the value of volume loss
for sands is not unique. In clays, it is correctly
assumed that volume loss does not change with depth
(undrained case, constant volume) and therefore calculation of volume loss derived from surface settlements
should give an accurate estimate of volume loss within
the tunnel. However, estimation of tunnel volume loss
using soil displacement data is not so simple for sands.
As Figure 4 illustrates, it depends very much on the
magnitude of tunnel volume loss achieved. The trend
of soil volume loss presented in Figure 4 was also
noted by Vorster (2005) and Jacobsz (2002) for similar
centrifuge tests in sands.
Examining Figure 4, we observe that the surface
and sub-surface volume loss are very similar up to a
tunnel volume loss of about 0.5%, after which the volume losses calculated at different depths within the soil
diverge. The maximum ratio of surface volume loss to
tunnel volume loss is 1.38 for the glass interface test
and 1.67 for the Perspex interface test. This maximum
ratio occurs at a tunnel volume loss of between 1.0
and 1.2%. As tunnel volume loss is increased further,
the rate of change of soil volume loss relative to tunnel volume loss begins to reduce. Surface volume loss
falls below tunnel volume loss at a tunnel volume loss
of between 2.3 and 2.6%. The slope for the subsurface
soil nearest to the tunnel is shown to increase to about
1 after a tunnel volume loss of approximately 3%.
Figure 4. Soil volume loss at various depths compared to

tunnel volume for test with a) glass interface, and b) Perspex
interface (VLsoil calculated using PIV displacement data).
Figure 5 presents a probable explanation for some

of the results shown in Figure 4 by examining volumetric strains within the soil (derived from the PIV
analysis of the Perspex interface test). Figure 5a) illustrates that at a low volume loss of 0.54%, the soil is
generally in a contractive mode in a fan-shaped zone
above the tunnel. This explains the tendency towards
larger soil volume losses at lower tunnel volume losses
in Figure 4. As tunnel volume loss is increased to 1%
and higher, a local zone of dilation develops in the soil
above the tunnel. As this dilatant zone grows in size
and magnitude, it compensates for the contractive soil
above it and results in lower calculated values of soil
volume loss, thus reducing the slope of the lines in
Figure 4.
PIPELINE ABOVE TUNNEL
Centrifuge testing of a model pipeline placed transversely above the tunnel was performed in order to
investigate soil displacements and structural behaviour
of the pipe.
705
Figure 5. Development of volumetric strain as tunnel volume loss is increased (from PIV data of Perspex interface
test).
The relevant dimensions of the test are given in

Figure 1. In prototype scale, the test represents a 4.65 m
diameter tunnel constructed at a depth of 13.65 m and
running transversely beneath a 1.43 m pipeline buried
at a depth of 5.25 m. The tunnel cover to diameter ratio
(CT /DT ) is 2.4 while the pipeline cover to diameter
ratio (Cp /Dp ) is 3.1.
Bending moment data was obtained using the displacement data of the pipeline provided by the PIV
analysis (Fig. 6). The vertical deflection data of the
pipeline was found to fit well to a modified Gaussian
curve (Equation 1) in the form presented by Vorster
et al. (2005).
where Sv = vertical displacement of pipe lining;

x = offset from tunnel centreline; Smax = maximum
vertical displacement of pipe lining; i = distance to
inflexion point of the displacement curve; n = shape
function parameter to control the width of the displacement curve; and = parameter to ensure i is consistent
with the Gaussian curve presented by Peck (1969). The
optimum fit was obtained using i = 55 mm and =
0.01 (resulting in n = 0.03). Note that when = 0.5
(n = 1), Eq. 1 becomes the Gaussian curve.
Bending moments were derived using elastic beam
theory whereby moments are related to the deformed
shape of a beam by:
Figure 6. Derivation of bending moments from deformed

shape of pipeline (from PIV data).
where M = bending moment; Ep =Youngs modulus

of the pipe material; and Ip = second moment of
the cross-sectional area of the pipe. Note that these
curves represent an approximation of the true bending
moments since the fitted curve can not match exactly
the actual deformed shape of the pipe.
There is a concern that tunnelling may cause gaps
below certain overlying structures. Figure 7 presents
PIV displacement data to illustrate that a gap does
indeed form below pipelines affected by tunnelling.
Soil displacements directly below the pipeline are
compared to pipeline displacements at the tunnel centerline. The upper and lower pipe linings are shown
to displace the same amount (i.e. no cross-sectional
distortion of the pipe occurs). Gap formation is shown
to commence at a tunnel volume loss of between 1
and 2%. The length of the gap (along the pipe) varied
from 2.9 pipe diameters at low volume loss to 3.4 pipe
diameters at 10% tunnel volume loss.
Figure 7 also shows a plot of the maximum sagging
moment of the pipe versus tunnel volume loss. The
data suggests that bending moments increase substantially with the onset of tunnel volume loss, however
pipe bending behaviour does not appear to be very
sensitive to the formation and growth of the gap that
forms below the pipe. The height of the gap (i.e. the
separation between the soil and the lower pipe lining)
is shown to increase significantly at a volume loss
of about 6% with no equivalent response in the pipe
706
1%
2%
3%
4%
5%
6%
8%
9%
10%
(3.4)
64
60
(3.2)
(3.1)
(2.9)
56
(bracketed values indicate

number of pipe diameters)
52
0.0
The pipeline test illustrated that a gap forms below

the pipe at a volume loss of between 1 and 2%. The
length of the gap grew from 2.9 tunnel diameters at
low volume loss to 3.4 tunnel diameters at 10% tunnel
volume loss. Bending moments induced in the pipe
increase from the onset of tunnel volume loss and do
not appear to be sensitive to the growth of the gap
height.
0
represents vertical
separation between
soil and lower pipe
lining
0.4
7%
68
maximum sagging
bending moment
(from PIV data)
0.8
1
2
3
1.2
1.6
bending moment (Nm)
gap length (mm)
0%
the maximum ratio of surface volume loss to tunnel

volume loss was as high as 1.67 and occurred at a
tunnel volume loss of between 1 and 1.2%;
a local zone of dilation formed just above the tunnel
at a tunnel volume loss of about 1%.
The zone of dilation effectively reduces the soil
volume losses above it.
PIV patches on top and

bottom lining of pipe
PIV patch on soil

just below pipe
6
2.0
2.4
0%
pipe - upper lining

pipe - lower lining
soil directly below pipeline
maximum sagging bending moment
1%
2%
3%
4%
5%
6%
7%
REFERENCES
8%
9%
8
10%
tunnel volume loss
Figure 7. Gap formation below pipeline and increase

in maximum sagging moment as tunnel volume loss is
increased.
bending moment. The width of the gap does not grow

substantially and is the likely reason why the bending
moments do not respond to sudden increases in gap
height at the tunnel centreline.
5
CONCLUSIONS
Soil-structure interaction mechanisms are important

when evaluating the response of buried structures
to ground movements. The research described in
this paper has elucidated the soil-structure interaction mechanisms that occur when tunnelling beneath
buried pipelines.
The results presented illustrate some important
aspects of tunnelling within sandy ground (i.e. drained
material). These include:
soil volume loss will not always be the same as
tunnel volume loss;
the value of soil volume loss depends on the
magnitude of the tunnel volume loss;
soil volume loss calculated at the surface can be
greater or lower than tunnel volume loss;
Attewell, P.B., Yeates, J. & Selby, A. R. 1986. Soil movements

induced by tunnelling and their effects on pipelines and
structures. Blackie and Son Ltd, UK.
Jacobsz, S.W. 2002. The effects of tunnelling on piled
foundations. PhD Thesis, University of Cambridge.
Klar, A., Vorster, T. E. B., Soga, K. & Mair, R. J. 2005. Soil
Pipe Interaction due to tunnelling: Comparison between
Winkler and Elastic Continuum Solutions. Geotechnique,
55(6), 461466.
Klar, A. & Marshall, A.M. 2007. Shell versus beam representation of pipes in the evaluation of tunneling effects on
pipelines. accepted, Tunnelling and Underground Space
Technology.
Klar, A., Marshall, A.M., Soga, K. & Mair, R.J. 2007. Tunneling effects on jointed pipelines. accepted, Canadian
Geotechnical Journal.
Mair, R. J. &Taylor, R. N. 1997. BoredTunnelling in the urban
environment. Proceed. 14th International conference on
soil mechanics and foundation engineering. Hamburg:
Balkema. 4:23532385.
Peck, R.B. 1969. Deep excavation and tunnelling in soft
ground. Proceed. 7th International conference on soil
mechanics and foundation engineering. Mexico City.
266290.
Vorster, T. E. B. 2005. The effects of tunnelling on buried
pipes. PhD Thesis. Cambridge University.
Vorster,T. E. B., Klar,A., Soga, K. & Mair, R. J. 2005. Estimating the Effects of Tunneling on Existing Pipelines. Journal
of Geotechnical and Geoenvironmental Engineering, 131
(11), 13991410.
White, D. J., Take, W. A. & Bolton, M. D. 2003. Soil deformation measurement using particle image velocimetry (PIV)
and photogrammetry. Geotechnique, 53(7), 619631.
707
Ground movement and earth pressure due to circular tunneling: Model tests
and numerical simulations
H.M. Shahin, T. Nakai, F. Zhang, M. Kikumoto, Y. Tabata & E. Nakahara
Nagoya Institute of Technology, Nagoya, Japan
ABSTRACT: Two-dimensional (2D) model tests on tunnel excavation using a newly developed circular tunnel
apparatus are carried out. Numerical simulations are also conducted using finite element method under planestrain and drained conditions. In the finite element analyses, elastoplastic subloading tij model is used as a
constitutive model of the ground material. From the model tests it is revealed that in the case of the same volume
loss due to tunnel excavation, surface settlement and earth pressure around tunnel are significantly influenced
by the displacement applied at the tunnel crown for the same overburden. The volume loss is less significant
compare to the crown drift in the case of shallow tunneling. The numerical results show very good agreement
with the results of the model tests.
INTRODUCTION
Shallow tunneling is one of the essential methods

to make underground space in urban area. With the
ongoing demand of tunneling technology research
works on ground movements and earth pressures due
to tunnel excavation have been potentially increased.
The trap door apparatus has been used to investigate the mechanism of tunneling problems by many
researchers (Murayama and Matsuoka, 1971; Adachi
et al. 1994, Nakai et al., 1997; Shahin et al., 2004).
In our previous work, we have been carried out laboratory model tests using trap-door tunnel apparatus to
investigate the deformation mechanism and redistribution of stress surrounding the tunnel. To investigate
the deformation mechanism and earth pressure of the
ground more precisely, a new and more realistic tunnel
apparatus has been developed where the cross section
of the tunnel is circular. The apparatus can simulate
various excavation methods such as full face excavation, top and side drift and bench cut excavation. This
paper reports 2D model tests using the newly developed circular tunnel apparatus and numerical analyses
using the subloading tij model. This model can consider influence of intermediate principal stress on the
deformation and strength of soils, Dependence of the
direction of plastic flow on the stress paths, Influence
of density and/or confining pressure on the deformation and strength of soils. Model tests are performed
with different soil covers and for different excavation patterns.The deformation mechanism is described
focusing the ground movements and shear strain
patterns.
Figure 1. Schematic diagram of 2D tunnel apparatus.
DESCRIPTION OF 2D MODEL TEST
2.1 Apparatus of model test

Figure 1 shows a schematic diameter of 2D tunnel apparatus. Figure 2 represents a newly developed
model tunnel with circular cross section. It consists of a
shim at the center of the tunnel surrounded with 12 segments. The segments are strongly tightened all around
the shim with rubber band. One motor is attached with
the shim to pull it out in the horizontal direction. The
model tunnel is kept in space with a vertical shaft, and
can be moved in the vertical direction with another
motor. Therefore, the device consists of two motors one is for shrinking the tunnel and the other for moving the tunnel vertically to fix it at a chosen ground
709
Figure 2. Circular tunnel device.
Figure 3. Schematic explanation of excavation patterns.
2.2 Model ground and excavation patterns

depth. It is possible to make these motors work simultaneously and individually together with controlling
the speed of the motors. With the motor the shim is
pulled out gradually which changes the diameter of the
shim, consequently the segments move inward and the
diameter of the tunnel is reduced. Changing the shape
of the shim different kinds of excavation process, such
as full face excavation, top and side drift and bench
cut excavation can be reproduced with this apparatus.
The reduction of tunnel diameter and the amount of
radial shrinkage are obtained from a dial gauge reading which is determined from the calibration result.
The vertical movement (if requires to impose) of the
tunnel is also measured with another dial gauge. Therefore, the shrinkage of the tunnel can be attained in a
controlled manner, which can simulate the condition
of a real tunnel construction.
In the apparatus 12 load cells are used to measure earth pressure acting on the tunnel. The load
cells are attached with the block which is placed surrounding the segments of the tunnel. Each load cell
block is 2.35cm in width and 5.0 cm in length. The
blocks are tightly fastened with rubber band. Therefore, earth pressure can be obtained at 12 points on
the periphery of the tunnel at a time. Earth pressure
can also be obtained at other positions by rotating
the tunnel. However, in this case it will be required
to make the model ground once again. Including the
load cell bocks the total diameter of the model tunnel
is 10.0 cm. The circular tunnel device is placed on an
iron table that was used for the trap door tunnel apparatus (Nakai et al., 1997; Shahin et al., 2004). It has
10 moveable blocks above which the ground is made.
The reason of using this type of base is to adjust the
initial stress condition of the ground such a way that
the stress distribution becomes similar to the ground
without tunnel. The surface settlement of the ground
is measured using a laser type displacement transducer with an accuracy of 0.0 1mm and its position in
the horizontal direction is incurred with a supersonic
wave transducer. Photographs are taken during experiments which are later on used as input data for the
simulation of ground movements with Particle Image
Velocimetry.
Firstly, the tunnel device is set at a height of 10cm; the

height is measured from the bottom boundary to the
tunnel invert. Varying the distance between the tunnel
invert and the bottom boundary, several experiments
were conducted. It was found that this height (10 cm) is
free from the influence of the bottom boundary. After
setting the tunnel device, mass of aluminum rods, having diameters of 1.6 and 3.0 mm and mixed in a ratio of
3:2 in weight, is stacked up to a prescribed depth. The
unit weight of the aluminum rod mass is 20.4 kN/m3 ,
and the length is 5.0 cm. The initial ground is made
in such a way so that the earth pressure becomes similar to the earth pressure at rest adjusting the bottom
moveable blocks of the apparatus. Great care is taken to
make a uniform ground and not to apply any undesired
load in the ground.
In this study two types of excavation patterns are
considered. Pattern 1 (full face excavation) corresponds to the excavation where the center of the tunnel
is kept fixed and the diameter of the tunnel is reduced
applying shrinkage of 4 mm all around the tunnel as
shown in Figure 3. Pattern 2 represents the excavation pattern where the invert is kept fixed (top drift
excavation). This is obtained by descending the tunnel
during the application of shrinkage. Here, the same
amount of shrinkage (4 mm) is applied. However, as
the center is moved downward by 4mm the amount
of imposed displacement at the tunnel crown is 8 mm.
In the both excavation patterns the volume loss of the
ground is the same, which is equal to 15.36%. The
model tests have been conducted for four kinds of overburden ratio, D/B equals 0.5, 1.0, 2.0 and 3.0, where
D is the depth from the ground surface to the top of
the tunnel and B (10 cm) is the width of the tunnel. In
Figure 3, dr represents the amount of shrinkage in the
radial direction towards the center of the tunnel, and dc
indicates the amount of descended of the tunnel center.
NUMERICAL ANALYSES
Figure 4 shows the mesh used in the finite element

analyses. Isoparametric 4-noded elements are used in
the mesh. Both vertical sides of the mesh are free in the
710
Figure 4. FEM mesh (D/B=2.0).

Table 1.
Material parameters for aluminum rods.
N (eNC at p = 98 kPa
&q = 0 kPa)
RCS = (1 /3 )CS(comp.)
e
0.008
0.004
0.3
Same parameters as
Cam-clay model
1.8
0.2
1.2
Shape of yield surface
(same as Cam-clay
at = 1)
1300
Influence of density and
confining pressure)
Figure 5. Stress-strain-dilatancy relation.
vertical direction, and the bottom face is kept fixed. To

simulate the tunnel excavation, horizontal and vertical
displacements are applied to the nodes that correspond
to the tunnel block. Analyses are carried out with the
same conditions of the model tests. In the case of the
fixed invert excavation, both horizontal and vertical
displacements are applied to the nodes correspond to
the blocks of the tunnel in the experiment except the
bottom most blocks, where only horizontal displacements are applied to the corresponding nodes. The
displacements patterns have been decided from the
actual movements of the ground in the experiments.
Two-dimensional finite element anlyses are carried out
with FEMtij-2D using the subloading tij model (Nakai
and Hinokio, 2004). Model parameters for the aluminum rod mass are shown in Table 1. The parameters
are fundamentally the same as those of the Cam clay
model except the parameter a, which is responsible
for the influence of density and confining pressure.
The parameter represents the shape of yield surface.
The parameters can easily be obtained from traditional
laboratory tests. Figure 5 shows the results of the biaxial tests for the mass of aluminum rods used in the
model tests. The figure shows the positive and negative dilatancy of aluminum rod mass; and it is clear
that the strength and deformation behavior is very
similar to those of dense sand. The dotted lines rep
resent the numerical results for a confining pressure
of 1/100 times the confining pressure of experiments.
From the stress-strain behavior of the element tests
simulated with subloading tij model, it is noticed that
Figure 6. Observed surface settlement profiles.
this model can express the dependency of stiffness,

strength and dilatancy on the density as well as on the
confining pressure. The initial stresses, correspondent
to the geostatic (self-weight) condition, are assigned
to the ground in all numerical analyses.
RESULT AND DISCUSSION
4.1 Surface settlement

Figure 6 shows the observed troughs of surface settlement in the full surface excavation and top drift
excavation for the amount of shrinkage dr = 4 mm in
the case of D/B = 1.0 and 2.0. Figure 7 represents the
computed results corresponding to the observed ones.
The abscissa represents distance from the center of the
tunnel, while the vertical axis shows the amount of
surface settlement. For both patterns of excavation the
maximum surface settlement occur vertically above
the tunnel crown. Surface settlements become smaller
with the increase of the tunnel depth, but they extend
over a wider region similar to the previous research
conducted with trap door tunnel apparatus (Shahin
et al., 2004). The shape of the surface settlement
711
Figure 7. Computed surface settlement profiles.
profiles is the same for all the soil covers in the case
of full face excavation. For the same volume loss the
maximum surface settlement is larger in the case where
the invert is fixed than that of the full face excavation.
This is because the applied displacement at the crown
is 8 mm for the fixed invert and 4 mm for the full face
excavation though the volume loss is the same. Surface
settlement occurs locally for 8 mm applied displacement in the case of fixed invert, consequently the shape
of surface settlement profile varies with the soil cover
in this case. The tendency of larger surface settlement
for the fixed invert is more significant up to D/B = 2.0.
However, in the case of D/B = 3.0 (not shown here) the
difference of the surface settlement significant. From
these results it is revealed that for the same volume loss
surface settlement profiles vary with the excavation
patterns in the case of shallower tunneling. Therefore,
the surface settlement may not be properly estimated
using the method of volume loss (Mair et al., 1993)
for shallow tunneling. The results of numerical analyses show the same tendency of model tests not only in
shape but also in quantity.
Figure 8. Distribution of shear strain: center is fixed.
is developed from the tunnel invert and covered the

entire tunnel during tunnel excavation for the full surface excavation. From Figure 9 it is seen that shear
band develops from the side of the tunnel not from the
tunnel invert. In this case the length of the shear band is
longer than that of the full face excavation. The range
of the deformed region for the fixed invert is narrower
compare to the full face excavation. The different patterns of shearing strain due to the different types of
the tunnel excavation lead the change of the ground
behavior. Moreover, the shear strain of the numerical
analyses shows very good agreement with the results
of the model tests.
4.3 Earth pressure

4.2
Shear strains
The distribution of shear strain of the model tests

are obtained from the simulation of Particle Image
Velocimetry (PIV) technique. The PIV is originally
developed in the field of fluid mechanics (Adrain,
1991). In this paper, two images are divided into a
finite area; the average movement rate of the aluminum
rods of each area is being output as nodal displacement. The strain for one grid is calculated from these
displacements by using the shape function and the
Jacobian matrix that are used in finite element method
for displacement and strain relationship. Figures 8 and
9 show the distribution of shear strain for full face
excavation and fixed invert excavation, respectively,
in the case of D/B = 1.0 and 2.0 for dr = 4 mm. It is
seen in Figure 8 that the shear band of the ground
Figures 10 and 11 show the observed and computed

earth pressure distributions for D/B = 1.0 and 2.0 in
the case of full face excavation and fixed invert excavation, respectively. The plots are drawn in the 12 axes
corresponding to the radial direction of the 12 load
cells towards the center of the model tunnel. The figures represent the value of earth pressure in Pascal
corresponding to the amount of applied displacement
(amount of shrinkage). It is seen in Figure 10 that
earth pressure decreases all around the tunnel for the
full face excavation due to the arching effect. The
results appear to be in agreement with the results of
tunnel experiments performed by Murayama and Matsuoka, 1971; Adachi et al., 1994; Shahin et al. 2004.
As shear band develops surrounding the entire tunnel (Fig.8) the surrounding ground undergoes to the
712
Figure 9. Distribution of shear strain: invert is fixed.
Figure 11. Distribution of earth pressure: invert is fixed.
Figure 10. Distribution of earth pressure: center is fixed.
loosen state which reduces stresses in that place. It

is also noticed that the earth pressure decreases suddenly after applying shrinkage of the tunnel within
0.00 to 0.20 mm. Further shrinking the tunnel, earth
pressure decreases gradually at a lower rate up to a
certain extent after which the earth pressure becomes
almost constant. Sudden change in earth pressure is
due to soil arching, immediately after disturbing the
ground. For the fixed invert excavation earth pressure
distributions are different from the full surface excavation. In this case earth pressure decreases all around the
tunnel till dr = 1 mm, for further shrinkage of the tunnel it increases in the bottom part of the tunnel while it
remains almost same in the upper part of tunnel. It can
be explained with the shear strain distribution shown
in Figure 9. As the ground becomes loosen only in the
upper part of the tunnel after dr = 1 mm, the confining pressure in the bottom part increases, therefore, the
increase of earth pressure in the bottom part of tunnel
can be speculated. From the above discussions it can
be said that the distribution of earth pressure is highly
dependent on the excavation patterns.
Figure 12 and 13 illustrates the change of earth
pressure at load cells 4 and 9 against the amount of
shrinkage of the tunnel for different soil covers. Load
cell 4 is located in the vicinity of the tunnel crown,
and load cell 9 is in the part of the tunnel invert. This
figure confirms the sharp change of earth pressure
during tunnel excavation. In the case of fixed invert
excavation (Fig. 13), at the position of load cell 9 earth
pressures increases after around dr = 1 mm, which is
different from the results of the full face excavation
(Fig. 12) where earth pressure remain constant after
that amount of shrinkage.The phenomenon of the earth
pressure increase after the reduction to some extent
can be described as the change of arching effect due to
the non-linear and elastoplastic behavior of the ground
materials. It can not be described with a usual linear
elastic model. The results of numerical analyses are
in good agreement with the results of model tests.
Therefore, it can be said that a proper elastoplastic
713
Figure 14. Stress-strain relation of aluminum rods mass.
Figure 12. Earth pressure history: center is fixed.
Figure 15. Comparisons of surface settlement profiles.
Figure 13. Earth pressure history: invert is fixed.
constitutive model is required to predict earth pressure

around tunnel for lining design.
of the model test. The elastic analysis produces a wider

surface settlement profiles compare to the observed
one, the maximum surface settlement is smaller as
well. As there is no yield point in a liner elastic model
it can not express the deformation occurred locally. In
this analysis displacement is applied to simulate tunnel excavation, therefore, there is no relation of the
magnitude of the Youngs modulus in the shape of the
settlement trough except the value of Poissons ratio.
COMPARISONS WITH THE ELASTIC

ANALYSES
In this study numerical analyses with a linear elastic theory has been carried out to compare the results
with the elastoplastic analyses. This section describes
some typical results of the analyses. Youngs modulus
for the elastic analyses is calculated from the stressstrain relation (Fig. 14) of biaxial test performed in
laboratory for the mass of aluminum rods. The value
of E = 5500 kPa is chosen from the figure, and the
assumed value of Poissons ratio is 0.33 for the ground
of aluminum rods mass.
Figure 15 shows the surface settlement profiles of
the model test, elastoplastic analysis and elastic analysis for soil cover D/B = 1.0. It is seen in this figure that
elastoplastic analysis can precisely express the results
CONCLUSIONS
To investigate the deformation mechanism and earth

pressure of the ground, a new tunnel apparatus has
been developed. With this apparatus 2D model test and
elastoplastic finite element analyses have been carried
out. From the model tests and numerical analyses, the
following points can be concluded:
1. Surface settlement and earth pressure around tunnel are significantly influenced by the displacement
applied at the tunnel crown for the same overburden
and the same volume loss.
2. The volume loss is less significant compare to the
crown drift in the case of shallow tunneling.
3. The full face excavation produces wider range of the
deformation region compare to the top drift (fixed
invert) excavation.
714
4. The ground deformation mechanisms are different

for the same volume loss with different excavation
patterns.
5. The distribution of earth pressure is highly dependent on the excavation patterns.
The finite element analysis with subloading tij
model is a useful tool to predict earth pressure and
ground behavior during tunnel excavation.
REFERENCES
Adachi, T., Tamura, T., Kimura, M. & Aramaki, S. 1994.
Earth pressure distribution in trap door tests: Proc. of 29th
Japan National Conference of SMFE, 3, 19891992 (in
Japanese).
Adrain, R. J. 1991. Particle imaging techniques for experimental fluid mechanics: Ann. Rev. Fluid Mech. 23,
261304.
Murayama, S. and Matsuoka, H. 1971. Earth pressure on

tunnels in sandy ground: Proc. of JSCE, 187: 95108 (in
Japanese)
Mair, R.J., Taylor, R.N. & Bracegirdle, A. 1993. Subsurface
settlement profiles above tunnels in clays. Geotechnique
43, No.2, 315320.
Nakai, T., Xu, L. & Yamazaki, H. (1997): 3D and 2D model
tests and numerical analyses of settlements and earth pressure due to tunnel excavation, Soils and Foundations,
37(3), 3142.
Nakai,T., and Hinokio, M. 2004.A simple elastoplastic model
material parameters. Soils and Foundation. 44(2): 5370.
Shahin, H. M., Nakai, T., Hinokio, M., Kurimoto, T., and
Sada, T. (2004): Influence of surface loads and construction sequence on ground response due to tunneling, Soils
and Foundation, 44(2), 7184.
715
Analysis of pre-reinforced zone in tunnel considering the

time-dependent performance
K.I. Song, J. Kim & G.C. Cho
Department of Civil & Environmental Engineering, Korea Advanced Institute of Science and Technology,
Daejeon, South Korea
ABSTRACT: Auxiliary support systems such as the reinforced protective umbrella method are applied before
tunnel excavation to increase ground stiffness and to prevent the large deformation in soft ground and shallow
depth tunnelling. This study suggests a method to characterize the time-dependent behavior of pre-reinforced
zones around the tunnel using elastic waves and direct shear test. The results obtained from the laboratory tests
are applied to numerical simulations of a tunnel considering its construction sequences. According to numerical
analyses, the time-dependent tunnel stability is most critical in the initial installation part of pre-reinforced zone
and the portal of tunnel. However, time-dependent effect on tunnel behavior is not significant during construction
as long as a proper overlap length is applied. Finally, the suggested analysis method of combining experimental
and numerical procedures that consider the time-dependent effect on the pre-reinforced zone on the tunnel
behavior will provide reliable and practical design and analysis for tunnels in soft ground.
INTRODUCTION
Recently, instances of the construction of large underground structures in soil and soft rock layers are
increasing. Auxiliary support systems such as the reinforced protective umbrella method are applied before
or during tunnel excavation to ensure the stability of
the tunnel face, to increase ground stiffness and to
prevent large deformations in soft ground during shallow depth tunneling. These auxiliary supports change
the state of stress and derive the arching effect around
the tunnel. However, the effects of pre-reinforcement
on tunnel stability and waterproofing of large section tunnels in soft ground have yet to be clearly and
quantitatively defined.
In conventional tunneling, there are typically 1
to 2 days between one face and the next face after
the pre-reinforcement step. During this time interval,
it is known that changes of the material properties
caused by the effect of curing of the grouting material exist. However, 28 days of stiffness and strength
after construction are generally applied to the material
properties of the pre-reinforced zone in design stage
without considering the effect of the time-dependent
behavior of the injected grout material.
The present study suggests a method to characterize the time-dependent behavior of pre-reinforced
zones around a large section of tunnel in soft ground
using elastic waves. An experimental analysis was performed to characterize the time-dependent behavior
of the pre-reinforced zone. Direct shear tests were

performed at different time stages to obtain the timedependent strength parameters. In addition, elastic
wave velocities (i.e., Vp and Vs) were continuously measured using piezoelectric bender elements to
obtain the time-dependent stiffness parameter. Timedependent strength and stiffness parameters obtained
from laboratory tests were applied in the numerical
simulation of a large section tunnel in soft ground,
taking into account its construction sequence. The proposed analysis method, which combines experimental
and numerical procedures while considering the timedependent effect on the pre-reinforced zone on the
tunnel behavior, will provide a reliable and practical
design basis and means of analysis for large section
tunnels in soft ground.
2 TIME-DEPENDENT EXPERIMENTAL
ANALYSIS
2.1
Measurement of elastic wave velocity and shear

strength
As shown in Figure 1(a), S-wave and P-wave velocity

were measured in this study using bender elements.
The propagation direction was varied by arranging the
installation direction of the bender element such that it
was parallel for the p-wave and perpendicular for the
s-wave. Wave velocities were measured continuously
717
Velocity (m/s)
1400
1200
1000
800
600
400
200
0
P-wave
S-wave
10
15
20
Curing time (Days)
(a) Elastic wave velocity (Normal scale)
Velocity (m/s)
1400
1200
1000
800
600
400
200
0
0.001
25
30
P-wave
S-wave
6 hours
0.01
0.1
1
Curing time (Days)
10
100
(b) Elastic wave velocity (Log scale)
Figure 2. Time-dependent characteristics of elastic wave

velocities.
2.2 Results and analyses of experimental tests
Figure 1. Schematic drawing of test setup.
Table 1.
2.2.1 Time-dependent characteristics of wave

velocities
The results show that the wave velocity increases
exponentially according to curing time and becomes
almost constant after 7 days. Figure 2(a) shows typical
velocity data for two types of measured waves. It is
found that the strength of the mixture becomes an
important factor from about 6 hours. As shown in Figure 2(a), P-wave velocity is faster than S-wave velocity
and represents the wave that travels through the skeleton of sand and cement mixture. From Figure 2(b), it is
seen that the wave velocity starts to increase when the
curing time is 46 hours. This means that the mixture
takes 46 hours to reveal a cementation effect.
Material properties fo samples used in test.
Material properties
Joomunjin
Standard Sand
Portland Cement
Friction angle ( )
Min. void ratio
Max. void ratio
Density (kN/m2 )
39
0.62
1.13
25.9
30.8
Realtive density (%)

Vertical Stress
(kPa)
60 W/C ratio (%)

160 Dimension
(cm)
170
10(W) 10(D)
7(H)
for 28 days under 160 kPa vertical stress. From

the wave propagation transition, arrival time was
calculated. The wave velocities were calculated by
dividing the height of the specimen by the measured
arrival time (Dano, 2004; Khan, 2006).
A schematic drawing of the motorized direct shear
apparatus is shown in Figure 1(b). The shear speed
of the direct shear apparatus was maintained at
0.5 mm/min. A load cell of 2 ton capacity was installed
to calibrate the load during shearing. Measured vertical and horizontal displacements were saved in a
computer automatically using a LVDT. All material
properties of samples used in the tests are listed in
Table 1.
2.2.2 Time-dependent characteristics of shear

strength and strength parameters (c and )
The direct shear tests performed for specimens less
than 7 days of curing time correspond to the wave
velocity measurement. The results of the direct shear
test were very similar to those of the wave velocity measurements. However, the tests were performed
with intervals of a day or more, and thus the process
of the strength revelation through cementation could
not be observed.
Figure 3 shows the time dependent characteristics of
shear strength parameters obtained from direct shear
test. In Figure 3(a), the friction angle does not change
in accordance with curing time. On the other hand,
718
Figure 4. Normalized shear modulus of sandy soil (Kim

et al. 2004).
Sandy soil in natural condition has no cohesion in the

direct shear test. However, after 7 days, it has a cohesion value of about 2.17 MPa, which is similar to the
result obtained by Eq. (1).
Figure 3. Time-dependent characteristics of shear strength
parameters.
Figure 3(b) shows time-dependent characteristics of

cohesion. Through Figure 3(b), it is noticeable that
the cohesion increases with curing time and after a certain amount of curing time the cohesion converges. It
can be deduced that the bonding of cement causes this
increase in cohesion. Therefore, strength is controlled
by the normal stress (i.e., frictional) at the early stages
and then by cohesion (Schnaid et al., 2001). In the
numerical analysis, it is assumed that the friction angle
does not change and the cohesion increases as curing
time increases.
Ultimately, the cohesion controls the increase of the
shear strength, and cohesion depends on time in a similar manner as elastic wave velocity. Hence, Youngs
modulus and cohesion are selected as improvement
properties to reflect time-dependent characteristics in
order to facilitate accurate simulation via the numerical analysis program in this study.
Youngs modulus of the grout bulb can be presented

as a function of the weight density and design criteria
strength in the concrete design specification, as given
by Eq. (2).
Here, when the weight density is 18 kN/m2 and

the unconfined strength at a curing time of 7 days is
8 MPa, Youngs modulus is about 8300 MPa. The initial Youngs modulus (i.e., 45 Mpa) can be calculated
from equation (3)(6), this represents an increase of
184 times.
Regarding the characteristics of elastic wave propagation, the maximum shear modulus can be calculated
by Eq. (3) from the measured shear wave velocity.
2.3 Verification of testing results and acquirement

of time-dependent design properties
An empirical determination method of pre-reinforced
zones to investigate the stability in a tunnel is presented
and the results are compared with characteristics of
elastic wave velocity from laboratory test results.
As shown in Eq. (1), cohesion is related to the
unconfined strength and friction angle. Herein, unconfined strength of grout bulb at 7 days of curing time is
about 8 MPa from the uniaxial compressive strength
tests. When the friction angle of a soil layer in natural
condition is about 39 degrees, the cohesion of the prereinforced zone is about 1.91 MPa according to Eq. (1).
The normalized shear modulus in the range of the

tunnel strain level can be obtained by a resonant column and torsional shear test and the Ramberg-Osgood
model (1943), as shown in Figure 4. Typically, the
strain level of a tunnel is about 0.11%. Hence, the
normalized shear modulus of sandy soil is about 0.3,
as shown in Figure 4. Accordingly, the shear modulus of the pre-reinforced zone can be calculated by the
maximum shear modulus from Eq. (3) and the normalized shear modulus in Eq. (4). Furthermore, the
Youngs modulus of model can be derived by Eq. (5)
719
and Eq. (6), which delineates the relationship among

Youngs modulus, shear modulus, and Poissons ratio.
Although it is clear that the measured shear wave

velocity has characteristics of infinitesimal deformation, it is reasonable that the increase in Youngs
modulus can vary according to any increase in the
elastic wave velocity depending on the time.
With these equations and theory, Youngs modulus of sandy soil in natural condition is 45 MPa
and Youngs modulus for a curing time of 7 days
is 7400 MPa. Therefore, Youngs modulus of a prereinforced zone increases 165 times compared to that
before pre-reinforcement. It is apparent from the equations (3) (6) presented here that stiffness increases at
a similar rate to the result calculated by Eq. (2). Moreover, the Youngs modulus of pre-reinforced sandy
soil after 28 days of curing time increases by about
200 times in comparison with normal condition sandy
soil. Based on these results and equations, numerical
analyses should be accompanied to evaluate the timedependent characteristics of pre-reinforced zones in a
tunnel.
Youngs modulus and cohesion of a pre-reinforced
zone should be expected to increase by 200 times
increase after a curing time of 28 days.
3
NUMERICAL SIMULATION OF LARGE

SECTION TUNNEL IN SOFT GROUND
Figure 5. 3D tunnel model and time-dependent material

properties of the pre-reinforced zone after 12 m excavation.
Figure 5 shows the simulated 3D four-lane tunnel

model, which was constructed 15 m under the surface
of the ground. It has a radius of 9.4 m and a height of
10.4 m. The boundary of 3D FE model is horizontally
fixed at the 4 end sections and vertically fixed at the
bottom section. Water table is located at the top of the
surface. This simulated tunnel is an actual section of a
four-lane tunnel that was designed and constructed in
weathered rock in Korea. Four sub-sectional excavations with a length of 0.75 m per stage were modeled.
Steel pipes 114.3 mm in diameter were modeled as a
beam element: a 12 m steel pipe was installed at an
inclination of 11 with a 6 m overlap, with a 0.7 m
transversal interval between the steel pipes that were
installed along the tunnel crown 60 from the center. The grout expansion radius is 0.3 m from the pipe
center.
3.2 Numerical analysis of a tunnel considering the
time-dependent behavior
3.1 Large section tunnel model and boundary

conditions
Comparative analyses were performed to analyze
the time-dependent effect of a pre-reinforced zone on
the large section tunnel behavior while tunneling. The
material properties of the pre-reinforced zone have
five different conditions that are either time-dependent
conditions or constant time conditions (i.e., 1d, 2d,
3d, and 28d). A commercial 3D FEM analysis program (i.e., MIDAS-GTS) was used as a numerical
simulation tool. Elasto-plastic ground material, which
is linear elastic and perfectly plastic with following
Mohr-Coulomb yield criterion and non associated flow
rule, was used for analyses.
For the application of the stiffness and strength
parameters depending on the time obtained from the
experimental study, a tunnel model should be located
at the same stress level of the experiment condition,
as propagation characteristics of elastic waves are
affected by the state of the stress.
The time-dependent behavior of a pre-reinforced zone

can be modeled using the procedures described below.
The material properties (i.e., stiffness and strength) of
the pre-reinforced zone are registered as the boundary conditions from Day 1 to Day 28. The registered
initial boundary conditions were applied to a preassigned mesh in the pre-reinforcement construction.
The boundary conditions were applied and updated
according to the field construction process. Figure 5
shows the conceptual time-dependent stiffness and
strength characteristics of the pre-reinforced zone after
the third pre-reinforcement and 12 m of excavation at
an excavation speed of 0.75 m/day.
The tunnel behaviors were analyzed considering the
time-dependent effect of the pre-reinforced zone in
terms of the vertical displacement and the horizontal displacement. The material properties used for the
analysis are tabulated in Table 2. Figure 6 shows the
time-dependent elastic modulus and cohesion values
obtained from the experimental study as well as those
used with the numerical analysis.
720
Table 2.
Ground
Material properties used for the numerical analysis.

(kN/m2 ) E (MPa)
Weathered soil 18
Weathered rock 21
Steel Pipe
33
45
200
210000
c (kPa) ( )
0.32 0
0.3 250
0.3
39
35
where : unit weight, E: elastic modulus, : Poissons ratio,

c : cohesion, : friction angle.
Figure 8. Normalized horizontal displacement variation at

tunnel face.
Figure 6. Time-dependent stiffness and strength for numerical analysis and obtained experimental result.
the analysis result of 23 days of constant time condition, and the vertical displacement converges. The
stiffness and strength of the pre-reinforced zone of
12 days of the constant time condition correspond
to 3050% of 28 days of the constant time condition.
In other words, a reduction of the material properties of
the pre-reinforced zone makes it possible to model the
time-dependent effect of the pre-reinforced zone on the
global tunnel behavior at the initial tunnel excavation.
4.2 Time-dependent effect on horizontal
displacement at the tunnel face
Figure 8 shows the normalized horizontal displacement at the tunnel face according to the timedependent condition and constant time conditions.
From the analysis result, variation of the normalized
displacement at the time-dependent condition ranges
from 0.940.98, which is very similar to other cases
during excavation. Therefore, pre-reinforcement can
be considered as prevention of a collapse rather than
a displacement reduction control at the tunnel face. It
can be concluded that grouting reduces the horizontal
displacement by approximately 26% at the tunnel
face with the pre-reinforcement method.
Figure 7. Normalized vertical displacement variation.
COMPARATIVE ANALYSIS OF THE

NUMERICAL SIMULATION RESULT
For quantitative analysis, displacements of each case

are normalized with the result of a pipe-only case.
With the normalized displacements, the effect of the
time-dependent behavior of the pre-reinforced zone
was examined.
4.1 Time-dependent effect on vertical displacement
Figure 7 shows the normalized vertical displacement
at the portal according to the time-dependent condition and constant time conditions. The normalized
vertical displacement behavior of the time-dependent
condition is similar to the results of one day of the
constant time conditions at the initial excavation section. As the excavation length increases, the results of
time-dependent condition become nearly identical to
4.3 Time-dependent effect on horizontal

displacement at the tunnel side wall
Figure 9 shows the horizontal displacement variation
at the tunnel side wall according to the time-dependent
condition and constant time conditions. From the analysis result, the normalized horizontal displacement
variation of the time-dependent condition is similar to
the results of one day of the constant time conditions
at the initial excavation section. The horizontal displacement variation of the time-dependent condition
is located between 2 and 3 days of the constant time
condition as the excavation length increases. Therefore, between 2 and 3 days of a constant stiffness and
strength condition can be used for a time-dependent
analysis of a pre-reinforced zone for the horizontal
721
Figure 9. Normalized horizontal displacement at tunnel

side wall.
and strength; secondly, the pre-reinforcement zone had

6 days of stiffness and strength.
The stiffness and strength levels of 6 and 14 days
of the constant time condition correspond to 87.8%
and 99.3%, respectively, of 28 days of the constant time condition. Therefore, the time-dependent
characteristics of the pre-reinforced zone do not affect
tunnel displacement due to the sufficiently overlapped
pre-reinforced zone. However, the time-dependent
characteristics of the pre-reinforced zone should be
considered in the case of a portal and an initial support
section of a weak layer, which can cause a large displacement and unsafe construction conditions. Therefore, it can be concluded that 23 days for the stiffness
and strength of pre-reinforced zones is appropriate
to model the time-dependent behavior. In addition, a
properly overlapped section of pre-reinforced zone can
be assumed as 3 or 7 days of constant stiffness and
strength without a time-dependent effect due to sufficient exposure of the stiffness and strength during the
excavation stages in conservative designs.
5
Figure 10. Normalized displacement at 10.5 m from the

portal.
displacement of a tunnel, which is identical to the

variation of the vertical displacement.
Figure 10 shows the normalized vertical and horizontal displacements 10.5 m from the portal. Figures 10(a) and 10(b) are in good agreement with the
result of 28 day of the constant time condition as well
as the result of the time-dependent condition of the
normalized vertical and normalized horizontal displacement. When the excavation length is 10.5 m, the
pre-reinforced zone initially has 14 days of stiffness
CONCLUSIONS
Based on an experimental study of a pre-reinforced

zone, the time-dependent characteristics were analyzed in terms of the strength and stiffness. Results
show that the shear strength and the elastic wave
velocities increase as the time stage increases. Shear
strength and strength parameters (i.e., the cohesion
and friction angle) can be uniquely correlated to
elastic wave velocities. The results obtained from
laboratory tests were applied to a numerical simulation of a tunnel, taking into account its construction
sequence. According to the results of the numerical simulation, vertical displacement and horizontal
displacement results for fewer than 23 days of constant time boundary conditions are nearly identical
to the analysis results of the time-dependent condition. Therefore, it can be concluded that 23 days
for the stiffness and strength of pre-reinforced zones
is appropriate to model the time-dependent behavior
of a large section tunnel. Finally, the suggested analysis method combining experimental and numerical
procedures that consider the time-dependent effect on
the pre-reinforced zone on tunnel behavior will provide a reliable and practical design basis and means of
analysis for tunnels in soft ground.
ACKNOWLEDGEMENTS
This paper was supported by the Korea Institute of
Construction and Transportation Technology Evaluation and Planning under the Ministry of Construction
and Transportation in Korea (Grant No. 04-C01) and
Brain Korea 21 Project in 2006.
722
REFERENCES
Dano, C., Hicher, P.Y. & Tailliez, S. 2004. Engineering
Properties of Grouted Sands. Journal of Geotech. and
Geoenviron. Engineering, 130(13): 328338.
Khan, Z., Majid, G., Cascante, G., Hutchinson, D.J. &
Pezeshkpour, P. 2006. Characterization of a cemented san
with the pulse-velocity method. Canadian Geotechnical
Journal, 43: 294309.
Kim, D.S., Kwon, K.C. & Oh, S.B. 2004. Development of
Experiment-Analysis Integrated System for the Evaluation of Deformation Behavior of Geotechnical Structures on Weathered Residual Soils at Whole Strain
Range. Research Report Korea Science and Engineering

Foundation.
MIDAS-GTS. 2005. Geotechnical & Tunnel Analysis
System, MIDAS Information Technology Co., Ltd.
Ramberg, W. & Osgood, W.R. 1943. Description of stressstrain curves by three parameters. Technical Note 902,
National Advisory Committee for Aeronautics, Washington, D.C.
Schnaid, F., Prietto, P.D.M. & Consoli, N.C. 2001. Characterization of Cemented Sand in Triaxial Compression. Journal of Geotech. and Geoenviron. Engineering, 127(17):
857868.
723
Vault temperature of vehicle fires in large cross-section road tunnel

K.S. Wang, X. Han & Z.X. Li
ABSTRACT: On the basis of CFD models, the simulation analysis of a lorry and car fire in large cross-section
road tunnel was carried out in this paper. The vault temperature was simulated and compared with the value
calculated by correlative empirical equation. The corresponding results would be contributed to provide effective
technique for the fire protection measures of the major structure of road tunnel.
INTRODUCTION
Large scale tunnel fires often caused considerable

material damage, not only to vehicles, but also to the
tunnel facilities as well.The damage was brought about
by the massive development of heat and aggressive
combustion gases, which led to enormous difficulties
to reconstruction of tunnel. In order to provide the road
tunnel with effective fire protection measures, it is
important to predict the vault temperature in different
road tunnel fire scenarios. There were some experiments on testing the structure characteristics under
fire. But most of them were tested by following the
standard temperature-time curve (Yasuda et al., 2004).
Very few data are available on maximum smoke temperature under the tunnel ceiling with different ceiling
heights, longitudinal ventilation velocities and fire
intensities. Based on CFD models, the simulation analysis of a lorry and car fire in large cross-section road
tunnel was carried out in this paper. The vault temperature was simulated and compared with the value
calculated by correlative empirical equation. The corresponding results would be contributed to provide
effective technique for the fire protection measures
of the major structure of road tunnel.
2
2.1
the conservation equations of mass, momentum and

energy are approximated as finite differences, and the
solution is updated in time on a three-dimensional,
rectilinear grid. Thermal radiation is computed using
a finite volume technique on the same grid as the flow
solver.
2.2
Simulation scenarios of tunnel fire
In this paper, CFD simulation of the maximum vault

temperature was conducted by FDS (4.06) in different
fire scenarios with constant longitudinal ventilation
velocity set as 3 m/s. The simulation length of the
model which was three-driveway large cross-section
road tunnel with diameter of 15 m was 100 m, as shown
in Figure 1. The Heat Release Rate (HRR) of the fire
DESIGN OF FIRE SCENARIO FOR CFD

Brief introduction of FDS
In these simulations, FDS (Fire Dynamics Simulator)

4.06 which was released by NIST (National Institute of Standards and Technology, USA) was used.
FDS is a Computational Fluid Dynamics (CFD) model
with LES (Large Eddy Simulation) of fire-driven fluid
flow. The model solves numerically a form of the
Navier-Stokes equations appropriate for low-speed,
thermally-driven flow with an emphasis on smoke and
heat transport from fires. The partial derivatives of
Figure 1. Cross-section of road tunnel.
725
25000
scenaio 1
scenaio 2
HRR (KW)
20000
15000
10000
5000
0
0
500
1000
TIME (S)
1500
2000
Figure 4. Heat release rate of fire source in scenario 1, 2.

Figure 2. CFD analysis model in scenario 1.
350.00
TEMPERATURE (C)
300.00
scenaio 1
scenaio 2
250.00
200.00
150.00
100.00
50.00
0.00
0
Figure 3. CFD analysis model in scenario 2.
source was based on the EUREKA project conclusion,

which was completed by nine Europe countries in antiquated road tunnel. The fire position was located at
37 m from the upstream ventilation cross-section and
63 m from the downstream ventilation cross-section.
In this simulation, the car was ignited by the lorry and
the design scenarios include: scenario 1: the car in the
flank of the lorry; scenario 2: the car downstream
the lorry. In these two scenarios, the spacing between
the car and the lorry was 2 m, as shown in Figures 2
and 3.
In the simulation process, the lorry firstly burned
and for about 220s the car was ignited by the lorry
which received more heat radiant flux than its critical
value of 16 kw/m2 . The maximum HRR of the lorry
and the car was 20 MW and 5 MW separately. The
total HHR generated by these two vehicles was shown
in Figure 4.
3
3.1
DESIGN OF FIRE SCENARIO FOR CFD

HRR curve
200
400
600
800 1000 1200 1400 1600 1800 2000

TIME (S)
Figure 5. Vault temperature-time curves in scenario 1, 2.
3.2 Vault temperature-time curve

As shown in Figure 5, the vault temperature reached
the maximum value of 321 C at the position of 4 m
downstream from the fire source after 941 s in scenario
1 and achieved 331 C at the position of 5 m downstream from the fire source after 980 s in scenario 2.
The vault temperaturetime curves were basically of
superposition in these two scenarios with the same
ventilation velocity of 3 m/s and geometry properties.
It was indicated that the relative position of these two
vehicles had less influence on vault temperature curve.
As shown in Figures 6 and 7, because of large crosssection of the road tunnel and relatively smaller HRR
value of the fire source, the simulated vault temperature was not much high. On the other hand, under
the longitudinal ventilation velocity of 3 m/s, the fire
flames were sloped towards downstream far from the
vault, so it attenuated the environmental temperature
and made less contribution to vault temperature.
3.3 Comparative analysis of vault temperature
As shown in Figure 4, HRR curves in these two scenarios were constructed. The relative position of the
lorry and the car had less influence on total HHR.
During the simulation, the HHR was approximately
23 MW and less than the algebra sum of 25 MW which
was calculated by the HHR of these two vehicles
respectively.
An empirical equation which was derived to predict

the maximum smoke temperature under the tunnel
ceiling is presented as follows (Kurioka et al., 2003):
726
Figure 6. Fire flames in scenario 1 after 940s.

Figure 8. Heat release rate of fire source in scenario 3, 4, 5.
where Q = heat release rate of the fire, kW; Q =

dimensionless heat release rate; Fr = dimensionless
Froude number; , = experimental constant; Cp =
specific heat capacity of air at constant pressure, kJ kg1 K1 ; a = ambient air density, kg m3 ;
Ta = ambient air temperature in tunnel, K; Tmax =
maximum excess temperature of smoke under the
tunnel ceiling, K; u = representative longitudinal ventilation velocity, m s1 ; g = acceleration due to gravity,
ms2 ; Hd = height from the surface of fire source to
tunnel ceiling, m.
For validating the availability of equation (1) to the
vault temperature in large cross-section road tunnel
fires, three other scenarios were introduced to the simulation analysis, including: (1) scenario 3: car fire;
(2) scenario 4: passenger car/lorry; (3) scenario 5:
heavy goods vehicle. The heat release rates of the fire
source were also in accordance with the EUREKA
project conclusion, as shown in Figure 8. The fire
flames in scenario 3, 4, 5 were shown in Figure 9, 10, 11
respectively.
The vault temperature reached the maximum value

at the position of 4 m downstream from the fire source
in scenario 3, 4, 5 with the same longitudinal ventilation velocity of 3 m/s. The simulation results and
empirical equation value of the vault temperature in
these scenarios were shown in Figure 12 and Table 1.
As shown in Figure 13 and Table 1, there were
rather differences between the simulation results and
727
the empirical equation values of the vault temperature. The relative error exceeded 20% for the HRR
which were less than 17 MW and greater than 68 MW.
It should be mentioned that for small scale fire, such as
HRR of 5 MW, the relative error approached 53.4%.
Therefore, much more simulation analysis and practical tests need to be carried out. On the other hand,
it was suggested that the empirical equation could be
further improved to be applicable for small scale fire
in large cross-section road tunnel. Perhaps some other
factors had to be considered during the calculation of
vault temperature, involving different kinds of ventilation conditions and geometrical properties of the road
tunnel.

simulation value
equation value
800
TEMPERATURE (C)
700
CONCLUSION
600
500
400
300
200
100
0
0
10
20
30
40
50
60
HRR (KW)
Figure 12. The simulation value and the equation value of

vault temperature in different scenarios.
Table 1. The simulation value and the equation value of
vault temperature in different scenarios.
3
4
1, 2
Scenario (5 MW) (20 MW) (23 MW)
Sv
Ev
Re
54
275
116
312
53.4% 11.9%
5
(50 MW)
(321 + 331)/2 = 326 753

369
628
11.7%
16.6%
1. As for lorry and car fires in large cross-section road

tunnel and under longitudinal ventilation velocity
of 3 m/s, the relative position of vehicles had less
influence on total HHR which was approximately
23 MW and less than the algebra sum of 25 MW.
2. Owing to large cross-section of the road tunnel,
small scale vehicle fire and longitudinal ventilation
velocity of 3 m/s, the simulated vault temperature
was not very high.
3. In connection with the calculation of the vault
temperature, there was limitation to the related
empirical equation. The relative error which was
between the simulation analysis results and the
empirical equation value exceeded 20% for the
HRR less than 17 MW and greater than 68 MW.
Obviously it should be further improved especially
for small-scale fire of the road tunnel.
ACKNOWLEDGMENTS
Sv = simulation value; Ev = equation value; Re = relative

error.
relative error
relative error
60.00%
REFERENCES
50.00%
Gu, Z.P. & Cheng, Y.P. 2003. Research on temperature field

distribution in transportation tunnel fires. Paper collections of first session international meeting on city and
industrial safety in China (Nan jing): 265268 (in Chinese).
Haack, A. 1998. Fire protection in traffic tunnels: general
aspects and results of the EUREKA project. Tunnelling
and Underground Space Technology 4(13): 377.
Kurioka, H., Oka,Y., Satoh, H. & Sugawa, S. 2003. Fire properties in near field of square fire source with longitudinal
ventilation in tunnels. Fire Safety 38 (4): 319340.
Yasuda, F., Ono, K. & Otsuka, T. 2004. Fire protection for
TBM shield tunnel lining. Tunnelling and Underground
Space Technology 19 (45): 317.
40.00%
30.00%
20.00%
10.00%
0.00%
10 17 20
30
40
50
HRR (KW)
60 68 70
80
90
Figure 13. The relative error among the simulation value

and the equation value of vault temperature in different
scenarios.
728
Effects of different bench length on the deformation of surrounding

rock by FEM
X.M. Wang, H.W. Huang & X.Y. Xie
Department of Geotechnical Engineering, Tongji University, Shanghai, P.R.China
ABSTRACT: Bench cut method has been extensively used in mountain tunneling. This is mainly due to
its flexibility to adapting to different ground conditions. Induced displacements are empirically controlled by
adjusting the speed of excavation, the bench length, partial-face excavation and closure of invert. In this paper,
a series of three-dimensional, numerical, elastoplastic analyses were conducted to investigate the effects of
different bench length on the deformation of surrounding mass in soft rock. The closure of invert was also
investigated to their role in controlling the final displacement. When bench cut method adopted in soft rock, the
bench length should not be too long or too short and 0.5 times of tunnel diameter around for bench length is
appropriate.
INTRODUCTION
In mountain tunneling, the bench cut method is used

extensively due to its simplicity and flexibility to
adapting to different ground conditions. And the
method provides an advantage of simultaneous excavation of the upper and lower sections. The key issue
of adopting this method is selecting a length and a
shape of bench to assure the stability of the face, especially for tunneling in soft rock. In addition, auxiliary
methods are used as required. When the ground is good
enough, having enough self-supporting properties, the
bench length can be reduced to 0 m (full face cut). In
the case of a weak ground, the bench length is empirically decided. It is therefore important to evaluate and
compare the effect of different bench length on the
deformation of surrounding rock and on the stability
of excavation front.
Usually, numerical analyses work as a kind of model
test in which many relevant design variables can be
investigated in parametric studies (Ng & Lee, 2005;
Karkus & Fowell, 2003; Galli et al., 2004). In this way
it is possible to quantify the relative importance of
each possible intervention in order to choose the most
effective measures from the economic and safety point
of view.
Seki et al. (1994) conducted a series of threedimensional finite-element elastic analyses of unlined
tunneling to determine the effect of bench length and
shape. The initial stress was given by external force,
ignoring the dead weight effect. They found that the

longer the bench, the smaller the displacement due
to squeezing at the face. Moreover, they found that
the bench length scarcely exerted influence upon settlement of the crown. Finally they concluded longer
bench tended to offer a great safety factor and leaving
the core was effective for increasing the face stability.
Farias et al. (2004) conducted a series of threedimensional elastic finite element analyses of tunneling by the New Austrian Tunneling Method (NATM)
to investigate relative importance of relevant techniques for settlement control. The techniques included
partial-face excavation, free span distance and support activation. The tunnel has 9.6 m of diameter and
the soil cover is 10 m. They found that tunnel support lining including free span and closure of invert
was the most relevant single factor analyzed in reducing induced settlements. The closer to face the lining
was concreted, the smaller the displacements. Moreover, they found the bench helped to keep horizontal
pressure in the excavation face.
Eberhardt (2001) conducted a series of comprehensive, three-dimensional, elastic and elasto-plastic,
numerical analyses of tunneling in the central Swiss
Alps with different assumed initial stress states to
demonstrate three-dimension stress rotation ahead of
an advancing tunnel face. The diameter of tunnel was
10 m and the bench length was 10 m. They found
that high stress concentrations in association with
large rotation of the maximum principle stress were
729
Table 1. Rock parameters used in the finite element

analyses.
Rock
E/GPa
/KN*m3
C/MPa
/( )
IVupper
IVlower
Vlower
6
2
1
0.3
0.35
0.35
23
20
17
0.7
0.2
0.05
39
27
20
Figure 1. Three dimensional finite element mesh.
observed when the initial maximum principle stress

alignment was horizontal and parallel to the tunnel
axis.
In this paper, a series of three-dimensional, elastoplastic, numerical analyses were carried out to investigate the effects of different bench lengths on the
deformation of surrounding mass in soft rock. The
objectives of these analyses are to find optimum bench
lengths for different rock mass.
2 THREE-DIMENSIONAL NUMERICAL
MODELLING
2.1
Numerical approximations
A hypothetical tunnel excavation in soft rock was modeled in this three-dimensional numerical study. The
diameter of the tunnel (D) was taken as 12 m, with a
constant cover depth 30 m. The tunnel was assumed to
be a bench excavation and lined with spray shotcrete
and bolt. The finite element program, MARC, was
adopted to model the tunnel excavation.
The three-dimensional finite element mesh used in
the present analyses is shown in Figure 1. The bedrock
was set 37.2 m below the bottom of the tunnel and the
domain expands laterally 80 m from the tunnel centerline. The model took advantage of symmetry of the
problem. Boundary conditions are totally fixed at the
bottom of the model and only vertical displacements
are free in the vertical sides. The model consisted of
10114 elements and 10373 nodes. Eight-noded brick
elements and four-noded shell elements were used to
model the rock and concrete lining, respectively. And
two-noded truss elements were used to model bolt.
The mesh was divided into 22 longitudinal blocks of
Figure 2. Schematic representation of L1 and L2 .
variable sizes. The first 20 blocks length corresponds

1/4 of the tunnel diameter (D). The last 2 block length
is 1/2 D. A monitored section, located in the middle of
the mesh (i.e., at z = 30 m), was studied during every
stage of excavation and construction. To account for
the relatively large stress and strain gradients near the
tunnel opening, small finite elements were used.
An elastic perfectly plastic rock model, using
Drucker Prager failure criterion with a nonassociated flow rule, was adopted in this study. The tunnel
lining and bolt were modeled as linear elastic. The
Youngs modulus and Poissons ratio for the tunnel lining were taken to be 25 GPa and 0.2, respectively. The
unit weight of the tunnel lining was 22 kN/m3 . For bolt,
theYoungs modulus of 210 GPa with Poissons ratio of
0.3, were adopted. Table 1 provides the rock parameter
used in this study. These values were based on Code
for design of Road Tunnel (2004).
2.2
Numerical modeling procedures
A given cross-section was divided into three parts:

upper section, lower section and invert section. For
simplicity, upper bench length equals to lower bench
length (Lb ). The free distance between excavation face
and the support heading will be referred as free span
(L1 ). The distance between the excavation face and
the first whole lining section will be referred as full
support distance (L2 ). The schematic representation of
these was shown in Figure 2.
730
6
Height from the springline (m)
4
0.25D
0.50D
0.75D
1.0D
1.5D
2.0D
top heading
full face
0
2
4
4
0.25D
0.50D
0.75D
1.0D
1.5D
2.0D
top heading
full face
0
2
4
0.05
0.00
0.05
0.10
0.15
Squeezing (mm)
0.20
0.25
0.30
0.0
Figure 3. Distribution of squeezing at the face (IVupper

rock).
0.2
0.6
0.8
1.0
Squeezing (mm)
1.2
1.4
Figure 4. Distribution of squeezing at the face (IVlower

rock).
Tunnel excavation and construction were simulated

by deactivating the rock elements within the proposed
tunnel excavation zone and by activating the support.
The tunnel excavation rate was modeled 3.0 m (i.e.,
D/4) per day, which was used as a step size in the
numerical analyses. No support was applied to the
tunnel face. The unsupported length equals 3 m (i.e.,
L1 = 3 m). The excavation sequences are:
0.25D
0.50D
0.75D
1.0D
1.5D
2.0D
top heading
full face
0
2
4
0
5
6
7
Squeezing (mm)
10
11
3.5
The initial stress was given by gravity because of

shallow tunnel. The different bench lengths (0-2D),
top heading cut and full face cut were studied in the
analyses.
3 ANALYTICAL RESULTS
3.0
IV lower
V lower
2.5
2.0
1.5
1.0
0.0
Squeezing at the tunnel face
When a large displacement is created because of

squeezing at the tunnel face, toppling or collapse by
slipping of the tunnel face is very likely to take place.
Figures 35 show, for each bench length, the distribution of squeezing (in the Z direction, i.e., tunnel driving
direction) along the tunnel centerline at the monitoring section. Figure 6 shows the relationship between
Figure 5. Distribution of squeezing at the face (Vlower rock).
Ratio of maximum squeezing to

that of top heading method
1 Excavate upper section rock until the upper bench

length equal Lb , and install tunnel lining to the previously excavation span simultaneously. Leave free
span of 3 m.
2 Excavate upper section rock and lower section rock
until the lower bench length equal Lb , and install
tunnel lining to the previously excavation span
simultaneously. Leave free span of 3 m.
3 Excavate upper section rock, lower section rock and
invert section rock, and apply lining.
4 Advance the tunnel by repeating step 3 until the
upper tunnel face has passed 3.0D from the monitoring section.
3.1
0.4
0.5
1.0
1.5
Bench length (D)
2.0
Figure 6. Relationship between bench length and maximum

squeezing displacement.
bench length (IVlower and Vlower rock) and maximum

squeezing displacement.
For different rock, as the bench length increases,
the squeezing distribution tends to the distribution of
731
Figure 7. Crown settlement versus face distance for different bench length (IVupper rock).
top heading cut. An increase in bench length greatly

decreases the maximum squeezing. The longer the
bench, the smaller the squeezing: for Vlower type rock,
compared with the results for the top heading method,
the squeezing of the full face method is 3.43 times,
with 0.25D bench length 1.49 times, with 0.50D bench
length 1.13 times, and with 0.75D bench length 1.03
times. For another two types of rock, as the bench
lengths increase, the changing trends of squeezing are
the same as the trend of Vlower type rock. However,
with the deterioration of rock, the stable value of bench
length (when the bench longer than stable value, there
is no significant change of squeezing) is different. For
the IVupper , IVlower and Vlower rock, the stable values of
bench length are 0.50D, 0.50D and 0.75D, respectively.
It is very interesting to note that for IVupper rock
the computed squeezing displacements at the top and
bottom of upper section are negative values. That is
due to load transfer in longitudinal direction resulted
from arcing in the unsupported zone. That has an
effect similar to the classical Terzaghis trap door
experiment.
3.2 Crown settlement
When a large displacement of the crown at the face
is produced, the tunnel face or its vicinity is prone
to failure. Figures 79 show the crown settlements
at the monitoring section versus the normalized distance of the excavation face to monitoring section,
for different bench lengths. A negative value of distance indicates that the excavation face has not reached
the monitoring section yet. A significant percentage of
the final stabilized settlement is induced before face
passage. This can only be adequately reproduced in
three-dimensional analyses.
The crown settlement before the face passage
does not significantly depend upon the bench length.
However, compared with full face method, the crown
Figure 8. Crown settlement versus face distance for different bench length (IVlower rock).
Figure 9. Crown settlement versus face distance for different bench length (Vlower rock).
settlement of bench method before the face passage has a remarkable decrease. As the bench length
increases, the stabilized settlement becomes larger and
closer to the value of settlement of top heading cut. For
Vlower rock, the stabilized settlement is 5.85 mm for
0.25D bench length, 6.41 mm for 0.50D bench length,
and 6.83 mm for 0.75D. When the bench length is
greater than 1.0D, the stabilized settlement value is
larger than 7.0 mm, but the exact value could not be
obtained in this study for the limitation of the model.
The reason for settlement decreasing mainly attributes
to full activation with invert closure. As the bench
length decreases, the value of L2 becomes smaller too.
From this point of view, to keep tunnel face stability,
the bench length should not be too long for soft rock.
On the other hand, to prevent tunnel squeezing, it is
necessary to make bench length long enough. Consequently, the bench length too long or too short is not
helpful to stability of excavation face, and there is a
reasonable value of bench length for some rock. For
732
good rock (i.e., IVupper rock), because the stabilized

settlement is small, it is not necessary to reduce bench
length to decrease the settlement and the bench length
can be set to one larger value which benefits the simultaneous excavation of the upper and lower sections.
For bad rock (i.e., IVlower and Vlower rock), the bench
length should not be too long, and 0.5 times of tunnel
diameter around for bench length is appropriate.
As mentioned in the introduction, Seki et al. (1994)
conducted three-dimensional finite-element analyses
to investigate the effect of bench length and shape
on the tunnel face stability. In their studies, they concluded that longer benches are made in poorer grounds.
However, in this study, the contrary result can be
obtained. This is because Sekis modeling was elastic
and the tunnel was unlined. The stabilized settlement
is the same for different bench length.
4
CONCLUSIONS
The bench cut method by which the upper and lower

sections are excavated at the same time is used extensively in mountain tunneling. Results were presented
from a detailed three-dimensional finite element analyses directed towards the effects of different bench
lengths on the deformation of surrounding mass in
soft rock.
The longer the bench, the smaller the displacement
due to squeezing at the face. When the bench length is
longer than stable value, there is no significant change
of squeezing, compared with the squeezing of top
heading method. The stable values of bench lengths
are different for different rock.
Compared with full face method, the crown settlement of bench method before the face passage has
a remarkable decrease, and it does not significantly
depend upon the bench length.
There is a reasonable value of bench length for some

rock. The bench length should not be too long or too
short and 0.5 times of tunnel diameter around for bench
length is appropriate for soft rock.
ACKNOWLEDGMENTS
This work was supported by Shanghai Leading Academic Discipline Project, Project Number: B308.
REFERENCES
Eberhardt, E. 2001. Numerical modeling of three-dimension
stress rotation ahead of an advancing tunnel face. International Journal of Rock Mechanics & Mining Sciences
38(4): 499518.
Farias, M.M., Junior, A.H., Assis, A.P. 2004. Displacement
control in tunnels excavated by the NATM: 3-D numerical
simulations. Tunneling and Underground Space Technology 19(3): 283293.
Galli, G., Grimaldi, A., Leonardi, A. 2004. Threedimensional modeling of tunnel excavation and lining.
Computers and Geotechnics 31(3): 171183.
Karakus, M. & Fowell, R.J. 2003. Effects of different tunnel face advance excavation on the settlement by FEM.
Tunneling and Underground Space Technology 18(5):
513523.
Ministry of Communications of the Peoples Republic of
China. 2004. Code for design of Road Tunnel. Beijing:
China Comunications Press.
Ng, C.W.W. & Lee, G.T.K. 2005. Three-dimensional ground
settlements and stress-transfer mechanisms due to openface tunneling. Canadian Geotechnical Journal 42(4):
10151029.
Seki, J., Noda, K., Washizawa, E. et al. 1994. Effect of bench
length on the stability of tunnel face. In Abdel Salam (ed.),
Tunnelling and Ground Conditions: 531542. Rotterdam:
Balkema.
733
The effects of loaded bored piles on existing tunnels

J. Yao, R.N. Taylor & A.M. McNamara
City University, London, UK
ABSTRACT: This paper presents the development of a series of centrifuge tests carried out to investigate the
effects of loading of bored piles on existing tunnels. The apparatus was designed to monitor the tunnel lining
deformation while a live loading was applied. Four facts were considered: rate of loading the pile, pile base level
relative to the existing tunnel, the ratio of the depth of clay cover to tunnel diameter, and the distance between
the pile and tunnel.
INTRODUCTION
High-rise buildings with deep pile foundations are

more and more used in the fast developing urban
environment. Inevitable disturbance to the ground and
surrounding underground structures caused by their
construction and subsequent loading may have significant impact in terms of settlement and deformation.
Over the last thirty to forty years, tunnel owners
have become concerned about this possibility, and
an exclusion zone was introduced to protect the tunnels. However, these guidelines, are mainly based
on their empirical correlations from similar projects,
and generally apply limits on the minimum distance
between the existing tunnels and new pile foundations.
(Chudleigh et al. 1999).
Many researchers have summarised the pile-tunnel
interaction problems (Schroeder, 2002;Yao et al 2006),
which can be categorised into two groups: effects of
tunnelling on piles and effects of piling on tunnels,
where the second group received less attention. Some
field case studies have been presented, which investigated the effect of piling on existing tunnels (Chapman
et al, 2001, Higgins et al, 1999, and Benton & Phillips,
1991). Schroeder (2002a, 2002b) also conducted a set
of FE analyses to investigate the interaction between
pile foundations and existing tunnels. Yao et al (2006)
described a series of centrifuge tests designed to investigate the effect of bored pile excavation on existing
tunnels.
The effect of bored pile foundations on existing tunnels can be categorised into two parts: pile installation
and the post piling period. In this project, the main
objective is focused on the post piling period, which
is the investigation on the effect of pile loading on
existing tunnels.
The paper presents the centrifuge apparatus design,

the centrifuge test results and some preliminary analysis of the results. All the tests presented are solely with
regards to post piling; the influence of pile excavation
is not considered.
2
CENTRIFUGE MODELLING
The apparatus was designed to meet five requirements:

the pile could be loaded at anytime, a range of loading
rates, the applied load could be recorded, pile settlement could be recorded and the rate of loading could
be changed during the test.
2.1
Model overview
Figure 1 shows a picture of the model on the centrifuge swing. The centrifuge strongbox is constructed
from aluminium with a transparent Perspex window
in the front to enable a view of the experiment in
progress. A tunnel lining with an internal deflection
gauge system installed in the centre of the sample, two
rub bags were used to seal the tunnel at the two ends
of the tunnel lining (Yao et al, 2006) and a pile loading
apparatus was fastened on top of the model container.
Druck PDCR81 pore pressure transducers were used
to monitor the pore water pressure in the model and
air pressure in the tunnel lining system during the centrifuge testing. Digital image processing was used to
trace the deformation at the front surface of the model.
All the tests were conducted in the Geotechnical Engineering Research Centre at City University,
London. The Acutronic 661 centrifuge with a maximum payload of 200 kg at 200 g and radius of 1.8 m is
described in detail by Schofield and Taylor, (1988).
All tests presented in this paper were conducted at
735
Figure 2. Schematic of the loading unit. C: depth of the

cover above tunnel, D: diameter of the tunnel.
Figure 1. Model on centrifuge swing prier to test.
an acceleration level of 100 g, according to scaling

law (Taylor, 1995), 1 cm in the model equals to 1m at
prototype scale. Kaolin clay was used for modelling
(Al Tabba, 1987).
2.2
Pile loading unit
A pile loading system was designed to load the preinstalled pile during the centrifuge test, and also
provide the facilities to measure the settlement of the
pile and applied load. Figure 2 shows a schematic of
the pile loading unit, which consist a model pile (preinstalled in the model), an actuator, two LVDTs, a load
cell, and support units.
For ease of centrifuge operation, the pile was
pre-installed into the model during model preparation. Having a pile pre-installed has two important
requirements:
No settlement would occur due to the self weight of
pile
The pile must be strong enough to withstand the
applied load
A 22 mm outer diameter, 10 mm internal diameter
aluminium tube with two ends sealed was selected as
the model pile. Two different lengths were made for
different pile base levels and cover to tunnel diameter
(C/D) ratios. The model pile was made to have a lower
density than the clay sample, so it would not cause
any settlements during centrifuge testing without any

applied load. As can be seen from Figure 2, a loading
pin sitting on top of the pile cap, which had two LVDTs
(Linearly variable differential transformers) installed
on either of it to monitor the settlement of the pile. The
entire loading unit will stop the pile being pushed out
due to the lighter density.
The actuator contains a 35 mm 90 watt motor,
RE35-118783, a planetary Gearhead GP 42 C gear
box, and an aluminium screw jack, MSZ-Alu. The
actuator fitted directly above the pile was supported
by a channel unit sitting on top of the strongbox as
can be seen from Figures 1 and 2. The loading pin was
attached to it, to apply the load to the pile. A load cell
was used to monitor the load applied to the pile. It was
mounted between the loading pin and the screw jack.
2.3 Deflection gauge tunnel system
Figure 3 shows the cross section of the model in plan,
which has similar layout to the pile excavation tests
carried out byYao et al (2006).The tunnel diameter was
50 mm with the thickness of 0.15 mm, and was made
of carbon fibre. A 16 mm diameter titanium tube was
used to support the tunnel unit and provide the pathway
for all the wires and cables. The tunnel was sealed
by two air-pressurised rubber bags at the two ends.
Tunnel lining deformation detector was attached to it.
736
Figure 3. Cross section of the model, showing the deflection

gauges and pile borehole(s). Dp is applied pile-tunnel spacing
during test (22 or 44 mm).
Figure 4. Assembled deflection gauges system.
Figure 4 shows the deflection detector unit developed

by Yao et al (2006). That was made of four deflection
gauges, which were calibrated strain gauged carbon
fibre cantilevers, and their output was recorded via
an onboard PC. In order to measure both inward and
outward deformation, the deflection gauges were given
a 4 mm pre-deformation.
2.4
Other equipment
Strongbox: The centrifuge model was set up within an

aluminium alloy strongbox, which had an inner plan
area of 550 mm 200 mm, and can contain a solid
model up to 300 mm high, a 80 mm width Perspex
window at the front to enable the model to be viewed
during the test.
Water was supplied from a stand-pipe sitting on the
centrifuge swing, and the pore pressure in the clay
model was monitored using pore pressure transducers
pre-installed into the sample before the test. Details of
this equipment have been well discussed by a number
of researchers at City University, London. (Taylor et al,
1998, McNamara, 2001).
A pile cutter set was used to excavate the bore hole
for the model pile. The cutter was made of 22 mm
stainless steel thin walled tube.
2.5 Sample preparation and test procedure
Kaolin slurry with a water content of 120% was poured
into the strongbox, which had a 3 mm thick porous
plastic sheet with a 0.5 mm thick filter paper in the
bottom. The strongbox was transferred into a consolidation press, and loaded up to a vertical effective stress
of 500 kPa. Upon completion of normal consolidation

it was swelled back to 250 kPa. The bulk unit weight
of the kaolin was about 17.44 kN/m3 . A de-aired and
calibrated pore pressure transducer was installed into
the sample after the swelling period.
A typical sample set-up and test procedure consisted
of the following steps:
Free water at the top of the model was removed after
closing the drainage taps closed at the base of the
strongbox; this was to avoid clay swelling back.
The applied vertical stress was reduced to zero
and the strongbox removed from the consolidation
press.
The front wall was removed, so the front surface
of the Kaolin sample could be cleaned to ensure a
better image process, and the top of the sample was
trimmed to the required height.
Tunnel was cut and pre-assembled tunnel unit was
installed into the model.
Marker beads for image processing were pushed
into the sample front surface on a 10 mm grid.
The Perspex window was then bolted onto the
strongbox.
The pile shaft hole was excavated and pile was preinstalled into the model.
The pile loading unit was mounted on top of the
strongbox; loading pin was driven down to the pile
cap.
The strongbox was weighed and placed on centrifuge swing.
The model was accelerated to 100 g on the centrifuge, and left over night to reach the pore pressure
equilibrium.
Air pressure in the rubber bags at the ends of the
tunnel was reduced
Load applied to the pile using the actuator.
The tunnel lining deformations, pore pressure in the
model, load applied, pile settlements and air pressure
in the rubber bags, were monitored and data stored on
the computer in the control room. Global movement
around the tunnel was measured using the image processing system. However, the results of digital image
analysis will not be discussed in this paper.
3 TEST RESULTS AND DISCUSSION
Eight tests were carried out, of which 2 were trial tests
used to commission and improve the apparatus. Table 1
summarises all the tests:
Pile was installed at two different tunnel-pile clear
spacing: 22 mm and 44 mm.
Two different C/D ratios were used: 2 and 3.
Pile base was installed at two different levels relative
to the tunnel position: tunnel crown level and invert
level.
Only a single pile was tested.
737
Table 1. Table of tests.
160
C/D
Pile Length
Pile base
Spacing
JY13
JY14
JY15
JY16
JY17
JY18
JY19
JY20
2
2
2
2
2
2
3
3
100 mm
100 mm
150 mm
100 mm
150 mm
100 mm
150 mm
150 mm
Crown
Crown
Invert
Crown
Invert
Crown
Crown
Crown
2.2
2.2
4.4
2.2
2.2
4.4
2.2
4.4
140
Shear strength
Test ID
120
100
80
60
40
20
0
20
*Spacing is the minimum distance between the outsides of

the tunnel and pile (m in prototype scale).
25
30
JY14
35
Depth (mm)
JY15
JY17
40
45
50
JY19
Figure 5. Undrained shear strength measured against depth

in model.
All the data presented in the paper are intended to

demonstrate the design of this research and the basic
data obtained from the centrifuge test.
3.1 Rate of loading the pile
The main purpose of two initial trial tests was to testify
the loading unit and to determine the best rate at which
to load the pile.
Burland et al (1966) produced a simple method
to predict the load/settlement curve. It is assumed
that the curve is linear up to full mobilisation with
takes place at a settlement of about 0.5% of the shaft
diameter. Brown et al (2002) presented a series of Statnamic tests to investigate the influence of loading rate
on pile behaviour in clay. The pile bearing capacity
was found to be particularly sensitive to pile deformation rate. Dayal & Allen (1975) found the similar
response. Frischmann & Fleming (1962) stated that the
recorded settlement was largely elastic. All settlement
was assumed to be as a direct result of shear strains.
Skempton (1951) presented the similar result:
where: s = shaft settlement, ps = load applied to pile

shaft-soil interface, L = effective length of pile shaft,
Gave = mean shear modulus of soil along pile shaft,
rm = radius from pile at which strain becomes negligible (Randolph & Wroth, 1978) and ro = pile radius.
Whitaker & Cooke (1966) stated that when the settlement is about 0.5 per cent of the shaft diameter, the pile
shaft frictional resistance develops rapidly and linearly
with the settlement.
The speed of pile displacement also affects the
pore pressure at the base of the pile, which can be
understood as on increase in magnitude of excess pore
pressure with increasing penetration rate (Brown et al,
2002). The rate of loading used in the centrifuge tests

were designed to create undrained conditions. The
loading speed of 2 mm per minute was selected on
the basis of the trial tests.
The ultimate load of a pile can be defined as either
the load at which the pile settlement continues to
increase without further increase of resistance, or the
load at which pile settlement reaches 10% of the
pile base diameter (Fleming et al, 1992). For most
soil conditions, the second category is more likely to
be the controlling factor for end bearing resistance
(Burland et al, 1966). In this research, the ultimate
load is defined as the load which causes a settlement
of 10% of pile foundation base diameter. The ultimate load capacity of piles can be estimated in terms
of undrained strength (Su ), in this research measured
from quick undrained tests direct taken after the test,
and undrained pile shaft adhesion factor, which was
chosen as 0.6 in this paper. Figure 5 shows the measured undrained shear strength again depth directly
taken after four tests.
Figure 6 shows the recorded load applied in the test
JY16. Half of the maximum calculated load reached as
soon as the load applied, then increased more gradually
towards the maximum. The load only achieved 85%
of the designed ultimate load. Applying the factor of
safety of 1.5 to 2, the achieved load was acceptable for
our research purposes.
3.2 Tunnel lining deformations

Tables 2 and 3 summarise the maximum deformations
measured by the deflection gauges for tests JY13 to
JY20, and positive values indicate movement towards
the tunnel centre. During the pile excavation test carried out by Yao et al 2006, the tunnel crown was
subjected to the greatest deformations for all cases,
and moved towards the tunnel centre, but there was
less effect at tunnel invert. As it can be seen, while the
pile was under load, the crown was still affected the
738
450
Applied Load (/)
400
350
300
250
200
150
100
50
0
0
10
20
30
40
Time (s)
50
60
70
Figure 6. The recorded increasing of load against time.

Table 2. Summary of maximum tunnel lining deformations
recorded during pile loading period (mm at prototype scale).
ID
Crown
Left
Right
Invert
JY15
JY16
JY17
JY18
JY19
JY20
11
18
22
10
24
16
1
1
2
1
3
2
2
2
3
1
4
3
8
12
19
9
18
14
Figure 7. The tunnel lining deformations against the applied

load. In all cases, positive values indicate movement towards
tunnel centre, deformations at model scale, for test JY16.
Table 3. Summary of maximum deformations in percentage: deformation over tunnel diameter (/D).
ID
Crown
Left
Right
Invert
JY15
JY16
JY17
JY18
JY19
JY20
0.22
0.36
0.44
0.2
0.48
0.32
0.02
0.02
0.04
0.02
0.06
0.04
0.04
0.04
0.06
0.02
0.08
0.06
0.16
0.24
0.38
0.18
0.36
0.28
Figure 8. Applied loads against pile settlements, for test

JY16.
Table 4. Change of tunnel diameter and tunnel centre
position (in prototype).
most, followed in turn by the invert, the right side and

the left side of the tunnel. (The right side of the lining
is the nearest site to the pile). Both right and left sides
were always moved away from the pile.
Figure 7 shows the recorded tunnel lining deformation for test JY16. It can be seen that the tunnel lining
did not start to move until the load reached half of its
maximum. Together with Figure 6, it shows the relationship between the applied load, pile settlement and
tunnel lining deformation. Similar results were found
for most of the tests.
Figure 8 shows the relationship between applied
load and pile settlements. The pile did not start to move
until the load achieved half of its maximum.
Table 4 lists the change of diameter of the tunnel lining in the vertical and horizontal directions. Figure 9
shows the relationship between the tunnel lining deformation and pile-tunnel clear spacing and Figure 10
ID
v/D
h/D
Cv
Ch
JY15
JY16
JY17
JY18
JY19
JY20
3
6
3
1
6
2
0.06
0.12
0.06
0.02
0.12
0.04
1
1
1
0
1
1
0.02
0.02
0.02
0
0.02
0.02
9.5
15
20.5
9.5
21
15
1.5
1.5
2.5
1
3.5
2.5
v and h : Change of tunnel diameter at vertical/ horizontal

direction (mm); Cv and Ch : Change of tunnel centre position
at vertical/ horizontal direction (mm), Cv move towards pile
opening and Ch move downward.
presents the relationship between the lining deformation and pile base position. It can be seen from changes
of the diameter, that deformation of the tunnel lining
is non-uniform. The changes of the tunnel centre position in the vertical and horizontal direction are also
summarised in Table 4. Figure 11 shows the change in
positions of the tunnel centre for each case. Both crown
and invert were subjected to significant impact for test
JY17 and JY19 where long pile with larger C/D ratio
739
Figure 11. The positions of tunnel centre for all cases.

Movement on X axis is towards pile opening, movement on
Y axis is towards the bottom of the strongbox.
Figure 9. The change of movement due to change of the

pile-tunnel clear spacing(C/D = 2).
Figure 10. The change of movement due to change of pile

base position (C/D = 2).
was used. Movements up to 0.48% of tunnel diameter were recorded on the crown when the long pile
was installed together with deeper cover. For the same
C/D ratio and tunnel-pile clear spacing, in test JY16
and JY17 a movement of 0.08% of the tunnel diameter was recorded for the longer pile. Summarizing the
tables and the figures, it can be see that in general,
as clear space increased, the deformation reduced; the
deeper the pile bases level or the longer the pile length,
the more the effect of the pile at the tunnel.
4
CONCLUSIONS
This paper presents a study of the influence of pile

loading on an existing tunnel, which is a part of a
research on the effect of bored pile installation and
subsequent loading on an existing tunnel. The design
of the model and preliminary analysis of the results
were presented.
Based on the literature review and the centrifuge
model tests, the following observations were made.
By reviewing previous research and based on two
trial tests, the rate of the loading was chosen as
2 mm per minute. The trial tests also confirmed that

increasing in the rate of loading will increase the
load achieved on the pile.
Pile settlement has a linear relationship with
increasing of the applied load once the load exceeds
half of the maximum designed ultimate load.
Tunnel lining response to the pile movement as
observed from the results shows that the tunnel
centre always move downwards and away from the
pile.
Increasing the pile-tunnel clear spacing will reduce
the deformation of the tunnel lining.
Using the long length pile will have more effect on
the tunnel lining regardless of the C/D ratio.
Tunnel crown is always subject to significant movement due to either pile excavation (Yao et al, 2006)
or pile loading. Tunnel invert affected more from
pile loading than pile excavation.
REFERENCES
Al-Tabbaa, A. 1987, Permeability and stress-strain response
of Speswhite kaolin, PhD thesis, University of Cambridge.
Benton, L. J. & Phillips, A., 1991, The Behaviour of two
tunnels beneath a building on piled foundations Proc. 10th
European Conference on Soil Mechanics and Foundation
Engineering, Florence, 665668.
Chapman, T., Nicholson, D. & Luby, D. 2001, Use of the
observational method for the construction of piles next to
tunnels. Proc. Int. Conf. Response of Buildings to Excavation Induced Ground Movements, (ed F. M. Jardine)
London:CIRIA.
Chudleigh, I., Higgins, K. G., St John, H. D., Potts, D. M. &
Schroeder, F. C. 1999. Piletunnel interaction problems.
Proc. Tunnel Construction & Piling 99, London. The
Hemming Group Ltd, 172185.
Higgins, K. G., Chudleigh, I., St John, H. D. & Potts
D. M. 1999, An example of pile tunnel interaction problems. Proc. Int. Symp. Geotech. Aspects of
Underground Construction in Soft Ground, IS-Tokyo
740
99(eds O.Kusakabe, K. Fuita & Y. Miyazake) Rotterdam:

Balkema, 99103.
Schroeder, F. C. 2002a, The influence of bored piles on existing tunnels: a case study. Ground Engineering 35, No 7,
3234.
Schroeder, F. C. 2002b, The influence of bored piles on existing tunnels. PhD thesis, Imperial college, University of
London.
Schofield, A. N. & Taylor, R. N. 1988, Development of standard geotechnical centrifuge operations. Centrifuge 88
(ed Corte), Balkema, 2932.
Taylor, R. N. 1995, Geotechnical Centrifuge Technology,

Blackie Academic and Professional, Glasgow.
Taylor, R. N. Grant R. J., Robson, S. & Kuwano, J.
1998, An image analysis system for determining plane
and 3-D displacements in soil models. Centrifuge
98, (eds, Kimura, Kusakabe & Takemura), Balkema,
7378.
Yao, J, Taylor, R. N. & McNamara, M. A. 2006, The effects of
bored pile installation on existing tunnels. The proceeding
of 6th International Conference on Physical Modelling in
Geotechnics, Hong Kong.
741
3D FEM analysis on ground displacement induced by curved

pipe-jacking construction
G.M. You
Shanghai Municipal Engineering Design General Institute, Shanghai, P.R. China
ABSTRACT: The power tunnel of Tibet road in Shanghai is the most difficult pipe-jacking project in China
with long distance and large diameter. The tunnel crosses through Suzhou river, underground pipelines and area
of dense buildings, especially the underground passage of subway and the existing subway line 2, so its vital
important to protect the surrounding environment when the pipe-jacking is constructed. The deformation of
ground surface and the existing tunnels during the construction of curved pipe-jacking is studied with 3D finite
element methods. First, the calculated results are analyzed and compared with the measured data in site to verify
the correctness of the 3D FEM model. Second, there are many factors which can affect ground displacement when
curved pipe-jacking is constructed. Among these factors, slurry sleeves qualities, soil pressure on the face of
pipe-jacking, slurry injection pressure and earth resistance are discussed on the basis of finite element simulation.
Then, the deformation of existing tunnels of subway line 2 is studied. Finally, the ground displacement formula
of curved pipe-jacking is discussed.The results show that the continuity of slurry sleeves and the pressure of the
face are very important factors to ground surface deformation during the construction of curved pipe-jacking.
The ground displacements induced by curved pipe-jacking are not larger than those of linear pipe-jacking if the
slurry sleeves around pipe are good and continual. Because of additional earth resistance, the maximum of the
ground surface deformation perpendicular to pipe axis is at the side of the center of the pipe-jacking curve, and
the distance of deviation depends on the radius of pipe-jacking curve.
1
INTRODUCTION
Curved pipe-jacking technology was applied in Japan,

Europe and America many years ago, and some successful experiences were also acquired. Nomura et al.
(1985) developed a pipe jacking method (D301) to
facilitate long-span, curved and high-speed capabilities in the construction of small diameter (300 mm)
tunnels. Nanno (1996) proposed a new curving method
called the unit curving method in which four jointadjusters are installed between pipes. The joint angle
is controlled by the adjusters and the thrust is also
distributed uniformly in the four adjusters. The new
method solved most of the technical problems in
curved drives and performed many jobs successfully
in actual construction sites. Vogler & Georg (2002)
studied the stresses on the curved pipe. The prediction equation for a curved jacking area was analyzed
in order to explain the characteristics of thrust and the
friction resistance (Shimada et al. 2004). The thrusts
in slurry pipe-jacking can be predicted accurately by
using the resistance between the mud slurry and the
concrete pipes and the resistance between the soil and
the pipes in the curved jacking area.
In China, pipe-jacking technology has been developed rapidly in recent years and many successful
constructions were obtained, such as the sewerage outfall project in Shenzhen (Mao, 2001), the secondary
project of improving combined sewerage system in
Shanghai (Ge, 2002), the project of Hefang street with
sewage pipe and rainwater pipe in Hangzhou (Jin et al.
2002), the project of main trunk sewer pipe of Yangli wastewater treatment plant in Fuzhou (Liu, 2003).
However, only single curve (horizontal curve or vertical curve) was used in these projects. Ding et al.
(2001) calculated and analyzed the jacking force, joint
stretching value, pipe internal force, stability of soil
and earth resistance of curved pipe-jacking by means
of pipe-joint mechanical model and beam on elastic
foundation method.
Fang & Wang (1998) analyzed the ground settlement due to pipe jacking in soft soils and developed
a method to predict the settlement profile of straight
pipe-jacking. Wei et al. (2003) analyzed the mechanism and reason of ground deformation caused by
pipe jacking construction. Wei et al. (2005) derived the
computing formulas of ground deformation induced
by bulkhead additive thrust, force of friction between
shield and soil, and force of friction between follow-up
pipes and soil by using the Mindlin solution in elastic
mechanics. Furthermore, the formula of total ground
deformation induced by pipe jacking construction was
743
obtained by combining the formula of ground deformation induced by ground loss with the previous one.
However, there is no study on the formula of ground
deformation of curved pipe-jacking.
The power tunnel of Tibet road in Shanghai is the
most difficult pipe-jacking project in China which is
3.033 kilometers long. This paper discusses the deformation of ground surface and the existing tunnels
by means of the numerical analysis. Some important
parameters during the construction and the ground
displacement formula of curved pipe-jacking are also
discussed.
2
2.1
Table 1. Physical and mechanical parameters of the

materials.
Materials
Unit
Youngs
Friction
weight modulus Poissons Cohesion angle
(kN/m3 ) (MPa)
ratio
(kPa)
(deg)
Brown clay
Silty clay
Muddy clay
Gray clay
Silty clay
18.6
17.3
16.9
17.3
17.8
26.7
14.2
11.8
16.1
27.2
0.35
0.37
0.39
0.36
0.35
21.6
14.4
14.0
15.7
18.5
14.5
13.0
10.5
11.5
17.5
3D NUMERICAL ANALYSIS OF CURVED

PIPE-JACKING
Mechanical model
The radius of pipe-jacking plane curve is 600 m. The

distance along the pipe axis range form 0 to 70 m.
The distance perpendicular to pipe axis is 66 m and the
height of the model is 40 m. The tunnel is located at a
depth of 7.5 m (from the top of pipe to ground surface).
The outer- and inner-diameter of the concrete pipe are
3,200 mm and 2,700 mm, respectively. In this analysis,
we use the following assumptions:
1. The pressure on the excavation face is uniformly
distributed with circular shape. The value is equal
to the actual measurement pressure of soil and
water cabin (face pressure). According to the actual
observation record, the face pressure is 0.18 MPa.
2. During the construction of pipe-jacking, time
dependent behavior of soil is not considered.
3. The frictional resistance between soil and pipe-ring
is uniformly distributed along pipe axis. According to the actual observation record, the frictional
resistance is about 2.0 kPa (You et al. 2006).
4. To distinguish with straight pipe-jacking, earth
resistance should be accounted which induced
by curved pipe-jacking. The relationship between
earth resistance and radius of pipe-jacking curvature is approximately inversely-proportional linear
(Ding et al. 2001). The earth resistance is 12.0 kPa
at a radius of 600 m.
2.2
Boundary conditions
Displacement boundary conditions are applied to this

model. The top side of the model is free boundary. Vertical displacements of the bottom side and normal displacements of the vertical sides are fixed, respectively.
2.3
Compute parameters
According to the geological report, the materials and

their parameters that are used for this simulation are
listed in table 1.
Figure 1. 3D-FEM mesh model.
The breaking criterion used in the model is DruckerPrager criterion.

2.4 Initial stress state
Initial stress state is obtained by the FEM software
directly. Only self-weight stress of soil is considered
without tectonic stress.
2.5 Finite element mesh
The mesh, consisting of 11120 nodes and of 10696
elements, is subdivided into 6 regions, having different
material properties. Eight node, solid element and four
node, shell element are used to simulate soil and pipejacking ring, respectively. In order to simulate the support effect of slurry sleeves, contact surface element
is engaged between pipe-jacking ring and outer soil.
Figure 1 shows the 3D FEM model used in this study.
2.6 Analysis of numerical results
Figures 2 and 3 show the comparison between ground
surface deformation obtained from numerical analysis
and form measured data. The predictions from the
FEM compare reasonably well with the observed
results.
744
Figure 4. Different locations of slurry sleeves around pipe.
Figure 2. Displacements of ground surface along pipe axis.
Figure 5. Displacements of ground surface perpendicular

to pipe axis with different locations of slurry sleeves.

to pipe axis.
Because of the complex factors affecting the deformation of ground surface, it is impossible to consider
every factor during the simulation of construction.
Only several main factors are involved in this study.
These factors are slurry sleeves, the pressure of the
face and the earth resistance.
2.6.1 Contribution of slurry sleeves
The effect of slurry sleeves to reduce resistance
between pipe and soil relates not only to the material
and mixture ratio of slurry but also to the injection parameters such as the location of injection
holes, injection pressure and slurry injection quantity.
The location of injection holes and slurry injection
quantity can be simulated by different locations of
slurry sleeves around pipe. In this study, 5 cases are
considered, as shown in Figure 4.
Figures 5 and 6 show the surface displacements with
different locations of slurry sleeves. Figure 5 indicates that the continuity of slurry sleeves has vital
important effect on the surface displacements. The surface displacements will be reduced sharply if good
slurry sleeves can be formed around the pipe. This
Figure 6. Displacements of ground surface along pipe axis

with different locations of slurry sleeves.
phenomenon is more obvious at the top center of pipe.

The range of deformation is wider if the slurry sleeves
are less continuous. Figure 6 indicates that the less
continuous the slurry sleeves, the lager is the ground
surface uplift movement ahead the tunnel face and the
less are the ground surface settlements above the tunnel face. The less continuous the slurry sleeves, the
less are the ground surface settlements behind the tunnel face (in the range 0 15 m). The settlements are
invariant after 15 m behind the tunnel face if the slurry
sleeves are continuous. Moreover, the settlements will
develop fast if the slurry sleeves are not continuous.
2.6.2 Contribution of face pressure
Figures 7 and 8 show the surface displacements with
different face pressures. Figure 7 indicates that face
745

to pipe axis with different face pressures.

to pipe axis with different earth resistances.

with different face pressures.
pressure has important effect on the surface displacements at the top center of pipe. Figure 8 indicates
that the larger the face pressure, the more obvious
is the uplift movement of surface ground. The point
with maximum uplift movement is closer to the tunnel
face with increase of face pressure. The deformations
ahead and behind the tunnel face will be increased
with increase of face pressure. The settlement at the
top of the tunnel face is invariant with different face
pressure.
2.6.3 Contribution of earth resistance
Figures 9 and 10 show the surface displacements
with different earth resistances. The alteration of earth
resistance does not change much the displacements
perpendicular to pipe axis. The displacements are
almost the same with different earth resistance. From
this result, it means that if good and continual slurry
sleeves can be obtained, the displacements of ground
surface of curved pipe-jacking are almost the same
as straight one. Moreover, because of additional earth
resistance, the deformation profile perpendicular to
pipe axis is not symmetric. The larger the earth resistance, the more obvious is the difference. Furthermore,

with different earth resistance.
the point with maximum ground surface displacement

perpendicular to pipe axis is not on the projection
line of pipe axis but at the side of the center of the
pipe-jacking curve.
3
NUMERICAL ANALYSIS OF SUBWAY

LINE 2S DEFORMATION
The power tunnel passed over the existing subway

line 2, and the minimum clear distance between power
tunnel and existing subway line 2 is 1.5 m. The total
projected length of crossover zone is about 25 m long.
The acute angle between power tunnel and subway
line 2 is about 75 degree.
Figure 12 shows the relationship between vertical
displacement of subway line 2 and jacking distance.
Figure 13 shows the relationship between horizontal
displacement of subway line 2 and jacking distance.
In Figures 12 and 13 origin of jacking distance is
40 m away from the cross point. For the total projected
length of crossover zone is about 25 m long, the jacking
746
Figure 13. Relationship between horizontal displacement

of subway line 2 and jacking distance.
Figure 11. Mesh of power tunnel and subway line 2.
Figure 14. Regression curve of ground surface displacements with 10 kPa earth resistance.
Figure 12. Relationship between vertical displacement of
subway line 2 and jacking distance.
distance ranged from 40 to 65 m when the power tunnel

passed through subway line 2.
Figure 11 shows the 3D finite element mesh of
the power tunnel and subway line 2. The region of
the model is: 100 m (along pipe axis) 60 m (perpendicular to pipe axis) 40 m (depth). Eight node,
solid element and four node, shell element are used
to simulate soil and pipe-jacking ring, respectively.
Displacement boundary conditions are applied to this
model. The top side of the model is free boundary.
Vertical displacements of the bottom side and normal
displacements of the vertical sides are fixed, respectively. The parameters that are used for this simulation
are listed in table 1.
In Figure 12 positive displacement indicates heaving movement. It can be seen that vertical displacement
of subway line 2 increased sharply after the jacking
face passed through the subway. There was almost no
displacement of subway line 2 before the pipe began
to traverse the subway. Because the pipelines traversed
above the subway line 2, only uplift displacements
occurred during the construction. Figure 13 indicates

that the direction of horizontal displacement pointed
to the tunnel face.
The FEM results show that the maximum of vertical and horizontal displacements were 3.3 mm and
0.39 mm, respectively. The values meet the demand of
the normal operation of existing subway line 2.
4
GROUND DISPLACEMENT EQUATION OF

CURVED PIPE-JACKING
Because of earth resistance, the deformation profile

perpendicular to pipe axis of curved pipe-jacking is
not symmetric to the normal line of pipe axis. According to numerical results as well as measured data,
the ground displacement formula of curved pipejacking is obtained by means of regression analysis
(Fig. 14). The vertical displacements perpendicular to
pipe axis are:
747
curved pipe-jacking are almost the same as straight

one. The key to control the displacements of ground
surface is to control the quality of slurry sleeves.
4 The face pressure is an important factor to ground
surface displacements. During the construction of
curved pipe-jacking, the pressure of soil and water
cabin should be controlled strictly.
5 Only uplift displacements of subway line 2
occurred during the construction of pipe-jacking.
The direction of horizontal displacements of subway line 2 pointed to the tunnel face.
REFERENCES
Figure 15. Relationship between curvature radius r and

regressive coefficient xc .
Where xc is the regression coefficient related to

the radius of pipe-jacking, Smax and i are the maximum
settlement of ground surface (when x = xc ) and standard deviation of the settlement curve, respectively.
Based on the research consequence (Ding et al.
2001), we can get the curvature radiuses correspondence to the earth resistances mentioned above
(10 kPa, 15 kPa and 20 kPa). Then, according to the
previous fitting result, the relationship between curvature radius r and regressive coefficient xc is obtained
(Fig. 15). The fitted linear regression line can be
expressed as follows:
It should be noted that the equations (1) and (2)

are obtained with the special geological conditions
of this project. The equations suitability for other
areas and geological conditions need more practical
verifications.
5
CONCLUSIONS
1 Because of additional earth resistance, the maximum value of the ground surface displacement
perpendicular to pipe axis is at the side of the center
of the pipe-jacking curve, and the distance of deviation depends on the radius of pipe-jacking curve.
The larger the radius of pipe-jacking plane curve,
the less is the deviation.
2 The deformation profile perpendicular to pipe axis
of curved pipe-jacking is not symmetric about the
normal line of pipe axis.
3 Slurry sleeves have vital effect on ground displacements. If good and continual slurry sleeves can be
obtained, the displacements of ground surface of
Ding, W.Q. & Zhu, H.H. et al. 2001. Beam on elastic foundation method considering pipe-joint and analysis of pipe
packing construction. Journal of Tongji University 29(5):
616620.(in Chinese)
Fang, C.Q. & Wang, C.D. 1998. An analysis and prediction of
ground settlement due to pipe jacking. Journal of Jiangsu
University of Science and Technology 19(4): 106110.(in
Chinese)
Ge, J.K. & Zhang,Y. 2002. Application of jacking technology
to sharply curved pile. Chinese Journal of Geotechnical
Engineering 24(2): 247250.(in Chinese)
Jin, W.H. & Gong, F.X. et al. 2002. Application of long distance curve pipe-jacking to Hangzhou project. Special
Structures 19(4): 6265.(in Chinese)
Liu, P.R. 2003. The Application of long-distance curved
pipe jacking technology. Fujian Construction Science and
Technology (2): 3839.(in Chinese)
Mao, B.Q. 2001. Technology of curved pipe-jacking with
long-distance and large-diameter in sea area. Special
Structures 18(3): 4853.(in Chinese)
Nanno, T. 1996. A method for driving curved pipe-jacked
tunnels. Tunnelling and Underground Space Technology
11(2): 325.
Nomura, Y. & Hoshina, H. et al. 1985a. Pipe jacking method
for long curve construction. Journal of Construction
Engineering and Management 111(2): 138148.
Nomura, Y. & Hoshina, H. et al. 1985b. Design and characteristics of D301 pipe jacking machine. Denki Tsushin
Kenkyusho Kenkyu Jitsuyoka Hokoku 34(12): 17891799.
Shimada, H. & Khazaei, S. et al. 2004. Small diameter tunnel
excavation method using slurry pipe-jacking. Geotechnical and Geological Engineering 22(2): 161186.
Vogler & Georg. 2002. Stresses on jacking pipes when
driving curved alignments. Betonwerk und Fertigteil
Technik/Concrete Precasting Plant and Technology 68(7):
5061.
Wei, G. & Huang, Z.Y. et al. 2005. Study on calculation
methods of ground deformation induced by pipe jacking construction. Chinese Journal of Rock Mechanics and
Engineering 24(Suppl. 2): 58085815.(in Chinese)
Wei, G. & Xu, R.Q. et al. 2003. Analysis of ground deformation caused by pipe jacking construction. Trenchless
Technology 20(6): 2427.(in Chinese)
You, G.M. & Ge, J.K. et al. 2006. Study on environmental
protection and control techniques of 3-D curved
pipe-jacking construction. Rock and Soil Mechanics
27(Suppl.): 398401.(in Chinese)
748
Theme 6: Calculation and design methods,

and predictive tools
Calculation of the three dimensional seismic stressed state of

Metro StationEscalatorOpen Line Tunnels system, which
is located in inclined stratified soft ground
R.B. Baimakhan, N.T. Danaev, A.R. Baimakhan, G.I. Salgaraeva, G.P. Rysbaeva,
Zh.K. Kulmaganbetova, S. Avdarsolkyzy & A.A. Makhanova
Scientific Center of Fundamental Research, Almaty, Kazakstan
S. Dashdorj
Hokkaido University, Sapporo, Japan
ABSTRACT: This includes examples of a destructive effect of earthquakes on the underground structures
and the new mechanicmathematical model of the inclined-stratified massif, which differs from the like model
in such a way that enables to research stressed state of an underground structure for arbitrary orientation of
the extended axis. Besides, this includes the results of calculating a three-dimensional seismic stressed state of
the underground system, which includes a metro station escalator open line tunnels in the soft ground that has
inclined and stratified structure.
INTRODUCTION
A lot of cases of damage and destruction of underground structures during a strong earthquake are
known. They include, for example, underground piping, mountain tunnels, stationary and metro open line
tunnels of both shallow and deep shaft. There are many
examples of considerable damage and strong destructions in underground structures, from piping to mines,
all around the world. Underground structures seem to
be stable. But this is not quite true. For example, as
shown on the Figure 1, in Daikai, Japan, during earthquakes in Kobe in 1995 (Magnitude 7.5) not only
the roofs and sides of the metro station but also its
columns were being destroyed laid at a epth of 50 m
from the surface (Iida et al. 1996, Ishihara 1998). There
occurred mass destruction in underground mines at
a depth of 326 m near the town of Shurab during
the Isfara-Batkent earthquake in 1977 (Middle Asia,
magnitude 6.5). (Rashidov et al. 1975).
Based on the analysis of damage of Based
on the analysis of damage of 71 underground
structures caused by earthquakes, Ch. Daudin,
D.V. Monakhenko, S.G. Shulman and others propose
to assess them by five categories: 1shifts along contacts; 2general distortion; 3local cracks; 4rock
fracturing and failure and 5partial failure (Monakhenko & Shulman 1980).
A general analysis shows that during intensive
earthquakes underground structures may be destroyed
Figure 1. Characteristic features of destructions of the

metro station during earthquakes in Kobe, Japan in 1995.
both in the shallow and deep shaft and also in the earth
stratum and rock mass. Factors, which effect on such
destructions are diverse. The strength of structure and
fastening elements depends not only on manufactured
materials, but also on physical & mechanical properties of the surrounding massif. However, one may note
a characteristic feature that underground structures of
deep shaft are destroyed owing to dynamic stresses
surpassing the breaking point of fastening materials.
751
The ground stratum surrounding an underground

structure often is characterized by natural anisotropy.
The existing methods allow us to consider static and
seismic stressed state mainly in a flat state. It is connected, firstly, with restricted possibilities of analytical
methods, and, secondly, with an undeveloped model
of underground structures in a three-dimensional state
subject to the arbitrary orientation of the extended axis
in space in arbitrary directions.
Thirdly, to fully provide seismic stability of a complex system of underground structures, it is necessary
to consider their interference (for instance, that of
station, escalator and metro paired open line tunnels
during seismic vibrations)
spreading. The longitudinal axis of an extended underground opening may have the same direction, i.e. the
opening may be inclined to the horizon. This variant
was first analyzed by R.B. Baimakhan (Baimakhan
2002).
He also has obtained the most general expressions
of velocities of quasi-longitudinal and two quasitransversal waves that spread in arbitrary direction in
the inclined layered transtropic massif with normal
n = n {cos , cos , cjs}.
For such massif, the equation of the generalized law
in a matrix form should be written as follows
MECHANICALMATHEMATIC MODELS OF
THE INCLINEDSTRATIFIED
TRANSTROPIC MASSIF
The Figure 2 includes different directions of wave

propagation in relation to the elements of the inclined
layered massif subject to the modeling by means of
a transversal isotropic body with an inclined plane of
isotropy. The variant I corresponds to the wave propagation along the line of layer spreading (intersection of
the inclined plane of isotropy and the horizontal plane
at the angle ), the variant IIacross the line of layer
spreading at the angle . They were previously considered by J.S.Yerzhanov, Sh.M.Aitaliev, J.K.Massanov in
the above-mentioned work.
Thereby underground openings are on the horizontal plane, though they are oriented in a different
way related to the line of layer spreading (plane of
isotropy): drifts and crosscuts, if they are oriented
towards the line of layer spreading across and perpendicular, respectively; diagonal openings if they are
oriented at an angle. The variant III corresponds to the
common case, where the direction of seismic waves
constitutes an arbitrary angle with the line of layer
The following values of elasticity factors cij for a horizontal stratified state are established. (Erzhanov et al.
1980).
where EK , K , (k = 1, 2) Jungs modules and Poissons ratios. G2 shear modulus. (Lekhnitskiy 1965)
Figure 2. Different directions of seismic waives spreading

in inclined-stratified transtropic massif relatively the lengthy
axis of underground structure.
where qim qjn , (p = , , , i, j = 1, 2, , 6) matrixes

of cosines of turning angles. As seen, said calculations
are interconnected. Based on the expression (5) in the
work of R.B. Baimakhan, expressions are received for
speeds of elastic wave propagation in the transtropic
752
massif with a normal of n = n {cos , cos , cos } in

an arbitrary direction as follows (Baimakhan 2002)
where = arccos (0.5q(p/3)3/2 ), environmental density, , , angles between the normal of

the wave front and the axes of the Cartesian coordinates
OXYZ.
3
the grounds along the Abay avenue between two rivers

Almatinka and Vesnovka. The Most values correspond
to grounds of flood-plain areas of rivers. Values of
elastic properties may be specified by way of detailed
determination of lithologic thicknesses of grounds in
regions of all stations and routes of driving tunnels.
If three components of the accelerogram on the surface are known, then to receive their modified values
with a depth along the vertical line down, according
to the work of R.B. Baimakhan, we have the following
formulas (Baimakhan 2002)
DESTRUCTIVE EARTHQUAKE
ACCELEROGRAM CONSTRUCTION
SUBJECT TO THE GROUND ANISOTROPY
The methods of reception of elastic equivalent characteristics for alternate isotropic strata are also available
in the work of J.S. Erzhanov, Sh.M. Aitaliev and
J. K. Masanov (Erzhanov et al. 1980).
For transtropic (transversal-isotropic) strata, elastic
constants should be calculated by using the following
formulas:
where hk thickness of the kstratum; n the number

of strata; E k , k , G k elastic characteristics of the k
isotropic stratum; 2(1 + k ).
With provision for data of JSC Almaty
metrokurylys and reference data on grounds (Bulychev 1989) some average values of elastic characteristics are systematised and adduced at in the Table 1.
Calculated by formulas (7) values of elastic constants
and for different locations of layers are adduced at the
Table 2.
These data by value of deformation modulus perpendicular to the layer E1 conditionally is possible to
refer to the three ground conditions of a city. Their
minimum values correspond to alluvial grounds at the
debris cones of rivers Big and Small Almatynka northward from the Raimbek avenue.Average meanings to
Table 1. Elastic properties of grounds E k , k and thickness

hk of their layers at the underground rote.
Ek
(m)
(m)
hk
Layers ground paper

1
2
3
4
5
6
7
8
9
10
7.0
30.0
25.0
8.5
120.0
80.0
50.0
22.0
200.0
300.0
0.40
0.36
0.28
0.21
0.27
0.35
0.25
0.36
0.32
0.31
2.8
2.2
3.3
2.7
3.2
4.8
6.5
4.5
5.1
5.9
Soft poured soil

Loamy soil
Gravel pebbles
Loam with pebbles
Boulder ground with gravel
Boulder ground with pebbles
Tick Loam
Sand of average size
Boulder with pebbles
Tick Clay
Table 2. Calculated equivalenttranstropic properties of

soft stratified ground E1 , E2 , G2 , 1 , 2 .
753
E1
(Mpa)
E2
(Mpa)
G2
(Mpa)
Layer
1
2
3
4
5
6
7
8
9
10
71.19
73.61
73.45
76.51
80.88
83.16
83.02
83.85
83.73
81.55
14.47
15.80
15.61
16.07
16.40
15.87
14.94
14.58
14.23
13.73
5.42
5.95
5.88
6.04
6.14
5.93
5.56
5.42
5.29
5.10
0.30
0.30
0.30
0.30
0.31
0.31
0.31
0.31
0.31
0.31
0.06
0.07
0.06
0.06
0.06
0.06
0.06
0.05
0.05
0.05
Figure 3. Synthetic accelerogram of a 10 Intensite

earthquake.
where q = 0.009;
Figure 4. Design area of a complex system of underground
structures.
Having divided the known values of the accelerogram on the surface of the ground by modification
factors, well receive their values a depth along the
vertical line down.
Using the formulas (2) (11) by the methods

mentioned in the work of R.B. Baimakhan, design
accelerograms of intensive 9 10 grade earthquakes
are constructed for a particular building site as related
both to the surface and the inner strata of a heterogeneous massif (Figure 3, Baimakhan 2003).
4
CALCULATION OF A SEISMIC STRESSED

STATE OF THE UNDERGROUND
STRUCTURE SPACE SYSTEM
By a method of finite elements, a seismic stressed state

of a complex system of the metro tunnels is researched.
The Figure 4 shows a design area. For clearness, the
metro station is shown in a double enlarged form.
Paired open line tunnels come to the metro station and go away from the opposite side. Boundary
conditions of the task are as follows: in the vertical planes ABDC and EFHG there are no horizontal
displacements.
= 0; in the vertical planes BFHD and AEGC there

are no horizontal displacements u = 0; in the horizontal plane CGHD there are no vertical displacements
w = 0. The upper horizontal plane AEFB is free of
stresses.The roofing of the metro station is at a depth of
17 m from the surface. Geometric dimensions: height,
width and length of said design 125, 45 and 160 m
respectively; those of the station 9, 22 and 100 m;
those of open line tunnels the height and width 2 m;
diameter of the escalator tunnel 9, length 54 m;
angle of inclination for the escalator against the horizontal axis 0X for the lying side 30 . Length, width
and height of the gallery, which connects the lower end
of the escalator to the station are 13, 8 and 5 m.
Said area is divided by eight-nods isoparametric
prismatic elements with eight additional internal integration points for 3508 space elements with total nods
of 4687. Physical & mathematical properties of the
massif are as follows: E1 = 1.028 104 Mna, 1 = 0.31,
2 = 0.10, E2 = 0.292 104 Mna, G2 = 0.11 104 Mna
= 2.2m/m3 . Lining materials for the station and the
open line escalator tunnels are:
E obdeku = 2, 5 104 Mna, obdeku = 0, 25,
obdeku = 2, 5 m/m3 (timberlining)
Values of the angles , and vary widely. According to the metro project, the angle of inclination
cannot exceed 4 for open line tunnels and 30 for
escalator and other special tunnels. The angles and
can change from 0 to 90 .
The Figure 5 shows seismic stresses diagrams at the
tunnel contours if = 0, = 0, = 4 and the angle of
wave fall = 60 , = 30 , = 0.
CONCLUSION
Under the static load, the areas of the roofing and the
lying side of the station are exposed to the main stress
concentration. Seismic stress diagrams, though they
754
are complex, however, show some regularities. In addition to the angular areas, side areas are exposed to the
most intensive seismic load. Such case is typical of
station, escalator and open line tunnels and galleries.
The middle part of the escalator tunnel is in a more
seismic stressed state.
The results of said research make it possible to
strengthen elements of an underground structure and
to reduce destructive effects cased by acts of God such
as earthquakes.
REFERENCES
Baimakhan, R.B. 2002. Analysis of seismic stability of underground structures in a heterogeneous thickness by a
method of finite elements. Almaty: Daur.
Baimakhan, R.B. 2003. Development of the analysis of seismic stressed state of underground structures subject to the
peculiarities of the geodynamics of the region. Thesis for
competition of an academic degree of Doctor of technical
sciences. Almaty: Institute Mathematical and mechanical.
Erzhanov, J.S. et. al. 1980. Seismic stability of underground
structures in the stratified anisotropic massif. Almaty:
Nauka (Science).
Iida, H. et al. 1996. Damadge to Daikai Subway Station. Soil
and Foundation. Special Issue: 283300.
Ishihara, K. 1998. Performance of tunnels and underground
structures during earthquakes. International Conference
on SoilStructure Interaction in Urban Civil Engineering.
89 October: 1931
Lekhnitskiy, S.G. 1965. Theory of elasticity of an anisotropic
body. Moscow: Gostechizdat.
Monakhenko, D.V. & Shulman, S.G. 1980. Issues of seismic
stability of underground structures. News from institutes
of higher education. Series Construction & Architecture
No. 8: 315.
Rashidov, T.R. et al. 1975. Seismic stability of metro tunnel
structures. Moscow: Transport.
Figure 5. Seismic stresses diagrams at the tunnel contours

in different times of a non-stationary seismic load of 10
grade earthquake: astation section at a distance of 22 m from
the front end; bstation section at a level of the escalator;
csection of open line tunnels at a distance of 35 m from the
station; dcentral section of the gallery between the escalator
and the station. The curves correspond: 1to the stati loading of the mass of the massif upper strata; 2 and 3 to the
seismic loads at the times of 18.244 sec. and 35.132 sec.
755
A complex variable solution for tunneling-induced ground

movements in clays
H.L. Bao, D.M. Zhang & H.W. Huang
ABSTRACT: A closed-form plane strain solution is presented for an elastic half-plane with a circular tunnel,
where an oval- shaped ground deformation pattern is imposed as the boundary condition of the displacement
around the tunnel opening. The solution translates complex variables from the research physical region onto a
circular ring in the mirror image region by conformal mapping. The coefficients in the Laurent series expansions
of the stress functions can be obtained from the boundary conditions and the convergence of the series. Finally,
a case study is also performed based on the metro line No. 2 in Shanghai and a satisfactory agreement between
the predicted settlement and the observed one is obtained.
INTRODUCTION
Due to recent city developments within the limited

land of urban areas, more and more complex facilities are developed under the ground surface, which
may cause serious potential damage to adjacent overlying services and structures. Engineers responsible
for the design and construction of tunnels should predict tunneling-induced ground movements in order to
protect the existing structures and tunnels. There we
three different approaches for estimating the potential tunneling-induced ground movements: empirical
methods, numerical methods and analytical methods.
Empirical procedures have been widely used to
assess potential ground movement owing to tunneling. In practice the ground deformations are often
described based upon field observations, for instance,
a normal Gaussian distribution curve proposed by Peck
(1969), which has no theoretical basis. It is assumed
that the surface settlement through can be approximated by the normal probability curve or error function. However in reality, ground movements depend
on a number of factors such as tunnel geometry and
depth, tunnel construction method, the quality of the
workmanship and management, behavior of the soil
around tunnel. Therefore the empirical methods are
also subject to these important limitations.
Recently some attempts have been made to develop
closed-form analytical solutions for tunneling-induced
ground movements in soft ground. Verruijt and Booker
(1996) presented a simple analytical solution for a tunnel in a homogeneous elastic half space by the virtual
image technique. Loganathan and Poulos (1998) introduced an equivalent undrained ground loss parameter,
which can be estimated using the gap parameter proposed by Lee et al. (1992). Verruijt (1997) proposed
the complex variable solution for circular tunnel in an
elastic half plane with the boundary condition of a prescribed uniform radial displacement at the cavity opening. Bobet (2001) presented another elastic solution for
ground deformations of a shallow tunnel in a saturated
ground which was made with the boundary condition
of uniform radial displacement at the tunnel opening.
However in practice the radial ground movement is
not uniform but oval-shaped. Park (2004) proposed
the elastic solution for the tunneling-induce ground
deformation in clay by imposing the prescribed four
different types of oval-shaped displacement at tunnel
opening. Wang (2007) expandedVerruijts solutions by
incorporating four different oval-shaped deformation
pattern at tunnel opening proposed by Park (2004).
In this paper, the complex variable method by Verruijt (1997) is also used for the solution of elasticity
problems for a half-plane with a tunnel of the third
oval-shaped deformation pattern at tunnel opening.
The metro line No. 2 in Shanghai is used to check
the applicability of the analytical solution.
2
STATEMENT OF THE PROBLEM
The problem deals with an elastic half-plane with a

circular tunnel, see Figure 1. The radius of the tunnel
is denoted by r, the depth of its centre by h, and the
757
Figure 1. Half-plane with a circular tunnel.
Figure 3. The region after conformal mapping.
The horizontal and vertical displacements, ux and uy ,

in the elastic region R can be expressed as
where is the shear modulus of the clay, and is

related to Poissons ratio by
As plane strain condition are assumed in this paper,

the stresses can be expressed as
Figure 2. Oval-shaped deformation pattern of tunnel

boundary.
cover by d. The upper boundary of the half-plane is free

of stress. As in practice the radial ground movement
around the tunnel is not uniform but oval-shaped. Park
(2004) proposed four different oval-shaped ground
deformation pattern as the inner boundary conditions.
Wang (2007) suggested the third boundary conditions
as the good agreement with the field measurements
had been obtained. Therefore in this paper, loading
takes place along the inner boundary of the tunnel,
in the form of the third oval-shaped deformation pattern (Figure 2), proposed by Park (2004), which can
be denoted by,
It is convenient to express the boundary condition in

terms of the integral of the surface traction, integrated
along the boundary,
where C is an integration constant.

4
BASIC EQUATIONS
The complex variable method for the solution of

two-dimensional linearly elastic problem involves two
analytic functions of complex variable,
(z) and (z).
CONFORMAL MAPPING
We conformally map the region R in the Z-plane onto

a ring region in the mirror image region, which
is referred as -plane, bounded by the circles || = 1
and || = , see Figure 3. The conformal mapping is
given by
758
boundary condition at the tunnel boundary, can be

written as a Fourier series,
Where
the coefficient must satisfy the equations
If 0 the radius of the circular caving is practically zero, which indicates a very deep tunnel. If
1 the covering depth is very small. For every
value of d/h the corresponding value of can be
determined from equation (8).
By virtue of the substitution z = () the function
(z) and (z) can be written in term of ,

and
Because the conformal transformation function

() is analytic in the ring region, the function
(z)
and (z) can be represented by Laurent series expansions,
Wang (2007) had given the expression of the coefficients of the Fourier series for the third oval-shaped
displacement pattern,
The coefficients of the series can be determined

from the boundary conditions.
5
BOUNDARY CONDITIONS
The first boundary condition is that the upper boundary y = 0 must be entirely free of stress. According to
equation (6), Verruijt (1997) gave the relationship of
the coefficients,
Therefore all the coefficients of the Laurent series

have been determined, except for a0 . This constant can
be determined from the requirement of convergence of
the Laurent series by the linear interpolation method
given by Verruijt (1997).
6
Therefore, one half of the unknown coefficients

have been expressed into the other half.
In this paper, the second boundary value problem
is considered, in which the third oval-shaped displacement pattern proposed by Park (2004) is prescribed
along the tunnel boundary. According to Verruijts
solution, if the function G (), which defines the
CASE STUDY
A case study is performed with the background of

Shanghai metro line 2. The tunnel lining for Shanghai
metro line 2 is 6.2 m in external diameter and 5.5 m in
internal diameter. The average depth to the center-line
of the tunnel is about 11 m. The tunnel was excavated
by EPB shield machine. The shield body is 6.24 m
long with 6.34 m in diameter. Therefore the clearance
between the external diameters of shield body and
the tunnel, which is usually named physical gap Gp
759
10
15
is not considered in this paper, which means that the

method in this paper can only be a preliminary design
of tunnels in soft clays. More factors should be taken
into account to improve the reliability of the method.
20
ground settlement(mm)
0
-10
-20
ACKNOWLEDGMENTS
-30
-40
The authors wish to acknowledge the financial support from National Science Foundation of
China (No. 50608058) and Hi-tech Research and
Development Program (863 Program) of China
(No. 2006AA11Z118).
Observed
Analytical
-50
-60
distance from tunnel center(m)
REFERENCES
Figure 4. Surface settlement.
according to the definition of Lee et al. 1992), will

reach 140 mm (Zhang,2004). The Youngs modulus of
the clay is taken as 25 Mpa and Poissons ratio as 0.3.
Figure 4 is a plot of comparison between predictions
of surface settlement from the complex variable solution and observed movements. The comparison is quite
good, which proves that the analytical solution can give
reasonable predictions of ground movement for shield
driven tunnels in soft clays.
7
CONCLUSIONS
Current design practice to predict tunneling-induced

ground movements is based primarily on empirical
methods, which are subject to some important limitations. In practice, the geometry of the tunnel, soil
properties and construction methods may affect the
settlements of the ground. In this paper, complex
variable method is used for the solution of elasticity
problems for a half-plane with a tunnel of oval-shaped
deformation. The metro line No. 2 in Shanghai is used
to check the applicability of the proposed analytical
solution. It gives good prediction of tunneling-induced
ground movements. However the plasticity of soft soil
Bobet, A. 2001. Analytical solutions for shallow tunnels

in saturated ground. Journal of Engineering Mechanics
127(12): 12581266.
Lee, K.M., Rowe, R.K. & Lo, K.Y. 1992. Subsidence due
to tunnelling: Estimating the gap parameter. Canadian
Geotechnical Journal 29(6): 929940.
for tunneling-induced ground movements in clays. Journal of Geotechnical and Geoenvironmental Engineering
124(9): 846856.
Park, K.H. 2004. Elastic solution for tunneling-induced
ground movements in clays. International Journal of
ground. 7th Int. Conf. on Soil Mech. and Found. Engrg.,
Mexico, 1969.
Verruijt, A. 1997. A complex variable solution for a deforming circular tunnel in an elastic half-plane. International
Journal for Numerical and Analytical Methods in Geomechanics 21: 7789.
Wang, L.Z. & Lv, X.J. 2007. A complex variable solution for
different kinds of oval deformation around circular tunnel
in an elastic half plane. Chinese Journal of Geotechnical
Zhang, D.M., Huang, H.W. & Hicher, P. 2004. Numerical
prediction of long-term settlements of tunnels in clays.
ITA-AITES 2004 World Tunnel Congress and 30th ITA
General Assembly., Singapore, May 2004.
760
Simulation of articulated shield behavior at sharp curve by

kinematic shield model
J. Chen, A. Matsumoto & M. Sugimoto
Nagaoka University of Technology, Nagaoka, Niigata, Japan
ABSTRACT: Tunnelling cases at sharp curve have increased by using articulated shield with copy cutter.
At sharp curve, the shield behavior, i.e., the position, the rotation angle, and the advance direction, should be
precisely controlled to follow the planed alignment. However, it is sometimes difficult to control and predict
the shield behavior without case records. To simulate the shield behavior, the kinematic shield model has been
proposed by the authors, taking into account the excavated area, the tail clearance, the rotation direction of cutter
face, the shield slide, and the dynamic equilibrium condition. This paper reports the simulation results of the
articulated shield behavior during excavation at a sharp curve. This study yields the following findings: 1) The
kinematic shield model simulates the measured path of the shield reasonably; 2) The articulated angle, the copy
cutter area and length are the predominant factors affecting the shield behavior.
INTRODUCTION
Recently, aiming at the completeness of city functions in urban areas, many infrastructures such as
roadways, water supply systems, sewerage and electric power lines, etc. have been constructed densely.
Due to spatial limitation, sometimes it is difficult to
choose simply an ideal route for a new tunnel. On the
other hand, with new techniques adopted continuously,
shield tunneling method has achieved greatly. Under
these circumstances, many shield tunneling cases at
sharp curve have been reported. In the case of excavation at a sharp curve, articulated shield becomes
popular since conventional single shield met some
difficulties in operational control and generated wide
range of ground disturbance.
The direction control systems have been applied.
However, these systems are based on empirical relationships and are lacking in the precise theoretical
background (Shimizu & Suzuki 1994). Therefore it
is sometimes difficult to control the shield at a sharp
curve and to predict the shield behavior without case
records.
To clarify the shield tunneling behavior and the
behavior of the surrounding ground, numerical methods such as FEM, DEM have been adopted (Komiya
et al. 1999; Kasper & Meschke 2004; Melis & Medina
2005). The enforced displacement by means of the
gap between the excavated surface and the tunnel lining has been applied in FEM and this technique gave
a very good prediction of the ground movement for
shield tunneling (Rowe & Lee 1992). DEM is proved
to be effective for soil stability problem at tunnel face.

However, these numerical methods require the shield
movement as one of known conditions. The immediate
ground movement during excavation stage is difficult to be simulated, since it is related to the shield
movement and the excavated area.
By taking into account ground displacement around
the shield, the kinematic shield model had been proposed for single circular shield (Sugimoto & Sramoon
2002). This model was validated by the simulation
of an earth pressure balanced (EPB) shield behavior
along a straight alignment in a single layer of sandy
gravel (Sramoon et al. 2002) and along a curve in a
multilayered ground (Sugimoto et al. 2007). Extending
this model to articulated shield, the articulated shield
model was developed and validated by simulating the
steady behavior of curve-only excavation (Sugimoto
et al. 2002).
The transient behavior of an articulated shield in a
way of sharp curvature movement is of interest to be
cleared. In this study, the simulation results of the articulated shield behavior at a sharp curve with radius 20 m
are reported. The validity of the model is examined by
comparing the simulation results with the measurement data. Furthermore, the factors affecting shield
behavior are also discussed.
2
KINEMATIC SHIELD MODEL
The articulated shield model is composed of five

forces: force due to self-weight of machine, f1 ; force
761
on the shield tail, f2 ; force due to jack thrust, f3 ; force

on the cutter disc, f4 ; and force on the shield periphery, f5 , as illustrated in Fig. 1 (Sugimoto & Sramoon
2002). Among them, f1 and f5 act on both sections of
the shield. f3 comes from shield jacks and articulated
jacks. f5 is composed of the ground reaction force and
the dynamic frictional force on the shield skin, which
are due to the earth pressure acting on the shield skin
plate. Since earth pressure is reliant on ground deformation, and excavated cross section area is usually a
little bit larger than the shield cross section area, the
ground reaction force from the ground to the shield
can be obtained by considering the coefficient of earth
pressure K, which is given by a function of the distance between the original excavated surface and the
shield skin plate Un , as shown in Fig. 2. In Fig. 2, K
at Un = 0 means the coefficient of earth pressure at
rest K0 , and the gradient of K at Un = 0 represents
the coefficient of subgrade reaction k. Here, note that
the subscripts h and v mean horizontal and vertical
directions, respectively; the subscripts min, o, and max
define minimum, initial, and maximum, respectively;
and v0 is overburden pressure. K in any direction
Kn can be calculated by interpolation between Kh
and Kv .
The shield behavior is represented by the movement
of the shield in x, y, and z directions (x, y, z), and
the shield postures (yawing angle y , pitching angle
p , and rolling angle r ). Since the change of r is
limited in practice, the factor of shear resistance due
to the cutter torque sg was adopted as a parameter
instead of r .
The articulated shield behaviour can be obtained by

solving the following equilibrium conditions of forces
and moments:
where F and M are the force and moment vectors

and the subscripts F and R denote front and rear sections of the shield respectively. The moment vectors
are generated by the cross product of the position
Figure 2. Ground reaction curve.
Figure 1. Model of loads acting on articulated shield.
762
vector and the force vector. Here, the superscript M

indicates the machine coordinate system, which can
be transformed to total coordinate system by using
transformation matrices.
SHIELD TUNNELLING SITE
3.1 Test site description

The test site was established at a cable tunnel. The
total length of this cable tunnel is 264 m and the test
site is about 60 m long at a leftward sharp curve section
with radius of 20 m. The inclined gradient of the tunnel
alignment is upward of 0.2%.
Figure 3 shows the geological profile at the test
section, where the overburden depth is about 34 m and
the groundwater level is 2.8 m below the ground surface. Table 1 shows ground properties. The ground is
composed mainly of alluvial layers (As1, Ac1, Acs,
As2, and Ac2) and diluvial layers (Ds1, Dc2, Dc3, and
Ds3). The tunnel is excavated in the diluvial sand layer

Ds3 with N -values over 50 by standard penetration
test.
The articulated slurry shield with outer diameter of
3.95 m and 5.765 m in length was used. After excavation, the reinforced concrete segments with outside
diameter of 3.8 m and 1.2 m in width were installed at
the straight sections. The steel segments with 0.3 m in
width were adopted at the curve section. The dimensions of the tunnel and the shield are summarized in
Table 2.
Table 2.
Dimension of tunnel and machine.
Item
Component
Value
Tunnel
Horizontal curve radius

(leftward)
Vertical slope gradient
(ascending)
Overburden depth
Groundwater level
Outer radius of segment
Width of segments
Outer radius
Total length
Length of front section
Length of rear section
Self-weight
Open ratio of cutter face
Thickness of cutter face
Radius of chamber
Length of chamber
Radius of cutter face
Number of jacks
Cross-sectional area
Radius of jack
Number of jacks
Cross-sectional area
Radius of jack
20 m
Shield
Shield jack
Articulated jack
Figure 3. Geological profile at test site.
Table 1.
0.2%
34 m
G.L. 2.8 m
1.9 m
1.2 m, 0.3 m
1.975 m
5.765 m
2.24 m
3.525 m
810 kN
20.0%
0.35 m
1.943 m
0.80 m
1.98 m
12
346.361 cm2
1.67 m
12
433.736 cm2
1.47 m
Soil parameters at test site.
Soil layer
Unit weight
(kN/m3 )
Cohesion
c (kN/m2 )
Friction angle
( )
Coef. of subgrade
reaction k (kN/m3 )
Coef. of earth
pressure at rest K0
Youngs modulus
E (kN/m2 )
Bs
As1
Ac1
Acs
As2
Ac2
Ds1
Dc2
Dc3
Ds3
19.0
19.0
15.3
18.1
19.2
15.4
19.2
15.8
15.0
19.7
0
0
59
0
0
50
0
91
160
0
26
38
0
24
29
0
37
0
0
45
4737
9474
789
3158
7895
2368
19737
8684
7895
67895
0.562
0.384
0.844
0.593
0.515
0.832
0.398
0.848
0.847
0.293
8400
16800
1400
5600
14000
4200
35000
15400
14000
120400
763
3.2
In situ data
Figure 4 shows the data of tunneling operation, shield

behavior, and excavation condition measured continuously by automatic measurement system. The parameters of tunneling operation are articulated angle in
horizontal direction CH (+: leftward), area of applied
copy cutter CC range (measured from the invert of
shield in clockwise direction, viewed from shield tail),
jack thrust F3r , horizontal jack moment M3p (+: rightward), vertical jack moment M3q (+: downward), and
cutter torque CT (+: anticlockwise direction, viewed
from shield tail). The parameters of observed shield
behavior are defined as yawing angle y (+: rightward), pitching angle p (+: downward), and rolling
angle r (+: clockwise direction, viewed from shield
tail). The parameters of excavation condition are shield
velocity vs , slurry pressure m , and slurry density m
in the chamber, which are usually controlled to stabilize the tunnel face. The mucking ratio Rv is the ratio of
the measured discharged soil volume to the theoretical
excavated soil volume.
The articulation of the shield was applied to negotiate the sharp curve. To follow the planed alignment, the applied articulated angle increased gradually
from the beginning point of the curve, then it kept
steady value of 500 minutes, which suits for the leftward curve of 20 m radius. CC was approximately
0.1 m in length and was applied mainly in range
of 30 180 (measured from the shield invert in
clockwise direction, viewed from tail) to increase the
excavated area around the cutter disc, which reduces
the acting earth pressure on the shield skin plate and
makes a shield advance easily. F3r was applied to drive
the shield forward against the earth pressure at the
cutter face and the dynamic friction around the shield
skin plate. After the transient section from curve to
straight line (from the distance 75 m to the distance
80 m), F3r increased obviously. M3p was applied to
turn the shield to follow the horizontal leftward curve.
M3q was mainly applied against the vertical moment
due to the earth pressure on the cutter disc and the
self-weight of the shield. CT was generated due to the
shearing resistance on the cutter disc.
The observed y reveals the detailed yawing behavior of the shield. The observed p indicates that the
shield negotiates the inclination of the tunnel alignment. The observed r fluctuates at the sharp curve
alignment and is somewhat relative to CT and the rotation direction of the cutter disc. This points out that the
shield rolls around its longitudinal axis in the opposite
rotation direction of the cutter disc.
The shield velocity vs decreased at the sharp curve.
To stabilize the face, m was applied based on the
lateral earth pressure at the tunnel face, and m was
kept approximately 12 kN/m3 . Rv was close to unity
throughout the test site, which indicates that excellent
excavation control has been achieved.
Figure 4. Shield tunneling measurement data.
764
is about 15 minutes uplift at the first straight section (from the distance 30 m to the distance 42 m),
compared with the observed data. From the distance
60 m to the end point of the test site, maximum 20
minutes uplift can be found. Considering the possibility of change of geological conditions at some
locations and the 5 minutes precision of the inclinometer, these differences are acceptable. Except for the
transient section from curve to straight line, the calculated vs is generally consistent with the observed.
At the transient section, the maximum difference of
0.01 m/min for vs is revealed. According to the construction report, intermittent excavation is applied to
confirm the tail clearance at this section, there is some
possibility to record slower velocity than that obtained
in the simulation.
4.2 Ground-shield interaction
Figure 5. Simulated and observed shield traces.
Figure 6. Simulated and observed shield behavior.
4
4.1
SIMULATION RESULTS
Shield behavior
Figure 5 shows the simulated and observed shield

traces. From this figure, the maximum difference of
1 cm for vertical position and the maximum difference of 3 cm for horizontal position can be verified.
The simulated and observed time dependent parameters y , p , vs are compared in Fig. 6. The simulated
y indicates that the shield performs good negotiation to the sharp curve. As for the simulated p , there
Ground-shield interaction is discussed by using the

calculated distance between the original excavated surface and the shield skin plate Un and the calculated
normal effective earth pressure acting on the shield ns
at straight line and sharp curve. Figs. 7 and 8 show the
distribution of Un and ns around the shield periphery
at the straight line with distance of 39.3 m and at the
sharp curve with distance of 58.4 m respectively. Here,
note that the shield periphery is unfolded as a flat plate,
i.e., the vertical axis shows the length of the shield and
the horizontal axis represents the circumference of the
shield.
In the case of excavation at straight line, the following are found from Fig. 7: (1) the contour lines of
Un become dense around 45 , 80 , 260 , 335 and Un
is about 30 mm around 60 and from 270 to 300 ,
because the copy cutter is applied from 300 to 90
shown in Fig. 4 and the effective rate of over cutting
decreased between 30 and 330 by the assumption
that the mucking at the invert is not sufficient; (2) the
distribution of Un at the cross section of the shield is
almost same along the longitudinal direction, which
is natural at straight line; (3) The contour lines of
ns appear around the invert and the crown of shield,
since Un around both spring lines is smaller than Un at
other area and the horizontal effective earth pressure is
smaller than the vertical one due to K0 of Ds3 = 0.293
shown in Table 1.
When shield excavated at sharp curve, the following are found from Fig. 8: (1) Un becomes positive
at the end of the front body and the rear body around
the right spring line of shield. At the same time, Un
becomes positive at the middle length of the both bodies along the left spring line. The shield skin plate at
these locations pushes the ground, whereas Un at the
opposite side becomes negative, where the earth pressure is in extension state. These characteristics result
from the equilibrium condition and correspond to the
leftward curve of the tunnel alignment; (2) the contour
765
Figure 7. Un and ns around shield at the straight line with distance of 39.3 m.
Figure 8. Un and ns around shield at the sharp curve with distance of 58.4 m.
lines of Un become dense at the right bottom of the rear

part and at the left bottom of the front part, since the
applying range of copy cutter shifts at the boundary
position due to the change of cutter face rotation direction shown in Fig. 4; (3) Un along both spring lines has
a fluctuation due to the wriggle motion of the shield
during excavation; (4) the intensity of ns appears at
the area where Un is positive, which is reasonable from
the view point of ground-skin plate interaction.
5
CONCLUSIONS
The articulated shield behavior at sharp curve was

simulated and the calculated shield behaviour was
compared with the observed one. Furthermore,

ground-shield interaction was discussed using the distribution of Un and ns around the shield. As a result,
the following conclusions can be made:
1. The kinematic shield model for articulated shield
simulated the shield position within 3 cm difference
and the shield pitching angle within 20 minutes
difference at sharp curve. This indicates that the
proposed model can simulate shield behavior at
sharp curve reasonably.
2. The articulated angle and the copy cutter area and
length are the predominant factors affecting the
shield behavior especially at sharp curve.
766
REFERENCES
Kasper, T. & Meschke, G. 2004. A 3D finite element simulation model for TBM tunneling in soft ground. International Journal for Numerical and Analytical Methods in
Komiya, K., Soga, K., Akagi, H., Hagiwara, T. &
Bolton, M. D. 1999. Finite element modelling of excavation and advancement processes of a shield tunnelling
machine. Soils and Foundations 39(3): 3752.
Melis, M. J. & Medina, L. E. 2005. Discrete numerical
model for analysis of earth pressure balance tunnel excavation. Journal of Geotechnical and Geoenvironmental
Engineering 131(10): 12341242.
Rowe, R. K. & Lee, K. M. 1992. Subsidence owing to tunnelling: II. Evaluation of prediction technique. Canadian
Shimizu, Y. & Suzuki, M. 1994. Movement characteristics and control method of shield tunnelling machine
of articulate type. Transactions of the Japan Society of

Mechanical Engineers (Series C) 60(571): 141148. (in
Japanese)
Sramoon, A., Sugimoto, M. & Kayukawa, K. 2002. Theoretical model of shield behavior during excavation II: Application. Journal of Geotechnical and Geoenvironmental
Sugimoto, M. & Sramoon, A. 2002. Theoretical model of
shield behavior during excavation I: Theory. Journal of
138155.
Sugimoto, M., Sramoon,A., Konishi, S. & Sato,Y. 2007. Simulation of shield tunnelling behavior along a curved alignment in a multilayered ground. Journal of Geotechnical
and Geoenvironmental Engineering 133(6): 684694.
Sugimoto, M., Sramoon, A., Shimizu, T., Dan, A. &
Kobayashi, T. 2002. Simulation of articulated shield
behavior by in-situ data based on kinematic shield model.
Journal of Tunnel Engineering 12: 471476. (in Japanese)
767
Deformation and pore pressure model of the saturated silty clay

around a subway tunnel
Z.D. Cui & Y.Q. Tang
X. Zhang
ABSTRACT: Shanghai subway Line No. 2 passes through the center of Shanghai from Songhong Road station
in the west to Zhangjianggaoke station in the east. The total length of Shanghai subway Line No. 2 is 25 km. Continuous dynamic monitoring is conducted by means of embedded earth pressure piezometers and pore piezometers around the tunnel. Based on continuous field monitored data and using the laboratory GDS (Global Digital
System) test apparatus, the developing law of the pore water pressure of the saturated silty clay around the tunnel is
explored with the distance. By the GDS test, the model of the increasing pore water pressure of the saturated silty
clay is put forward under the vibration loading. It is also amended by field monitored data.The pore pressure model
and characteristics of deformation of the saturated silty clay of Shanghai under the subway train are analyzed.
The result offers a valuable reference to the design, construction and the safe operation of the subway tunnel.
INTRODUCTION
The law of increasing pore water pressure under vibration loading is an important factor for the deformation
and strength of soil and is also the key to use effective
stress theory for the dynamic analysis. It is of primary
importance to predict correctly the changing law of
pore water pressure of soil. Researchers have studied
the increase and dissipation law of pore water pressure in sandy soil under vibration loading. Zeng et al.
(2005) and Guan et al. (2004) studied the characteristic of pore water pressure in silt and silty sand under
the cyclic loading and put forward the changing law of
pore water pressure. Zhou et al. (2002) and Pradhan
et al. (1998) tried to find the relation between the pore
water pressure and the path of strain and stress, and
summed up the effect on the development of the pore
water pressure in the liquefaction of the sandy soil.
Lee et al. (1975) and Shao et al. (2006) studied the
test parameters of deformation capability of the saturated sandy soil under the cyclic pore water pressure.
Li et al. (2005) analyzed the mechanism of liquefaction of silt, influence factors and the developing law
of pore water pressure of silt in the course of vibration. Guo et al. (2005) studied the effect of different
directions of the principal stress on the characteristic
of undrained circulation of saturated loose sand in the
course of vibration. Zeng et al. (2001) studied the

increase and dissipation law of saturated clay under
shock loading. Meng et al. (2004) studied the characteristic of dynamic response of pore water pressure in
saturated silty clay under shock loading. But there are
few researches on the developing law of pore water
pressure of saturated clay under the long term subway
vibration loading. The grain composition, mechanical
characteristics and moisture migration, etc. of the saturated silty clay are different from sandy soil. The speed
of the increase and dissipation of pore water pressure
under long-term subway vibration loading is relatively
slow and the developing law of pore water pressure is
also different from sandy soil. Based on continuous
field data and with the use of GDS in the laboratory,
this paper studies the law of the increasing pore water
pressure and deformation of the saturated silty clay
around the subway tunnel at different depths.
2
DYNAMIC MONITORING
In order to study the influence caused by the subway

vibration loading on the saturated silty clay around
the tunnel, the field test and monitoring are conducted
in this research. The site is selected between Jingansi
Station and Jiangsu Road Station. The dynamic monitoring system is adopted for field monitoring and its
769
sampling frequency can reach 200 Hz and its precision

is 0.1 kPa. It can fully reflect the soil response around
the tunnel due to the subway vibration loading. The
dynamic monitoring system consists of a resistance
sensor, a dynamic strain amplifier, a data selector and
a computer. The system can record all the sampling
data collected by the computer in real time.
Figure 1 shows the layout of boreholes at the site.
In the plane, there are five boreholes, each 110 mm in
diameter, parallel and vertical to the subway tunnel,
respectively. The distance between the site and Jingansi Station is 210 m. Boreholes No. 3, No. 4 and
No. 5 are parallel to the tunnel axis only 1.8 m away
from the outside of the segment of the subway tunnel
and the distance between them is 15.0 m. Boreholes
No. 1, No. 2, and No. 3 are vertical to the tunnel axis.
In order to study the attenuation of the effect on the soil
around the tunnel with the increasing distance under
the subway vibration loading, the distance between
No. 1 and No. 2 is 3 m and that between No. 2 and
No. 3 is 2 m, so that there is a step-up course. Figure 2
shows the distribution of strata and instruments. In the
section, the subway tunnel lies in gray silty clay of
layer No. 4. The earth pressure piezometers and pore
piezometers are located at the depths of 8.5 m, 11.5 m
and 13.5 m, respectively, in layer No. 4, to monitor the
response characteristic of the vibration for the subway
running.
3
to design the laboratory test. GDS (Global Digital

Systems) apparatus is used in the test, as shown in
Figure 3. It can monitor the test process at the real
time and collect data at high speed and store them. It
has merit of high precision, easy operation, reliable
results, etc. Its working principle is shown in Figure 4.
In the cyclic test, the soil samples are first saturated under back pressure. In order to simulate field
conditions to the utmost, the samples are consolidated under k0 condition. The confining pressure h
is obtained by calculation according
to the natural soil
strata, that is, h = k0 v (v =
i hi ). After consolidation, the cyclic triaxial test begins and the cyclic
stress d should simulate the bearing dynamic loading
to the utmost.
Strain gauge
2.70m
GDS TEST
Under the natural stress, the stress state of the undisturbed saturated silty clay is in the k0 consolidation.
The response amplitudes and response frequency of
soil have not been studied under the subway vibration loading. In order to study the characteristics of
soil under the subway vibration loading, this paper
uses field monitored data and field investigation data
1.30m
~ ~ ~ ~
~
~ ~ ~ ~
~ ~ ~ ~ ~~
~ ~ ~
~ ~ ~ ~
~ ~ ~ ~~
~ ~ ~ ~ ~
~ ~ ~ ~ ~
~ ~ ~ ~
~ ~ ~ ~
~ ~ ~ ~
18.0m
~ ~ ~ ~
~
~ ~ ~ ~
~ ~ ~ ~ ~~
~ ~ ~
~ ~ ~ ~
8.5m
~
1.8m
R3 m
11.5m~
~ ~ Tunnel
13.5m
~ ~
~ ~ ~ ~
~ ~ ~
~
20.0m
20.0m
7.80m
Figure 2. Soil profile and embedded instruments.

Note: No. 1 layer is mixed soils; No. 2 layer is brown yellow
silty clay; No. 3 layer is gray mucky silty clay; No. 4 layer is
gray mucky clay; No. 5 layer is gray silty clay.
Zhong Shan Park
No.3 Hole
No.4 Hole
2000
1800
Subway
People Square
Monitoring
Room
No.5 Hole
No.2 Hole
3000
No.1 Hole
15000
Figure 1. Layout of boreholes (unit: mm).
15000
Figure 3. GDS system for dynamic multi-function

triaxial test.
770
By the continuous dynamic field monitoring, plenty

of data of boreholes are obtained. By collecting, analyzing and arranging them, two kinds of response
frequency of soil are obtained when the subway train
runs across the site. The high frequency is 2.4 2.6 Hz
and the low frequency is 0.4 0.6 Hz. The test uses
the frequency of 2.5 Hz.
Moreover, the stress response amplitudes of soil at
different depths under the subway loading are obtained
by field monitoring. The maximum changing amplitudes of soil stress response are 0.23 kPa at 8.5 m,
0.70 kPa at 11.5 m and 1.15 kPa at 13.5 m. The change
of soil stress amplitude is approximately linear with
depth. Moreover, the stress amplitude in the rush hour
in the morning is larger than that in the evening which
is also larger than that at noon. The test adopts the
maximum stress amplitudes as reference for the worst
case in construction.
The field monitored data indicate that the soil
dynamic response around the tunnel under the railway
vibration loading is the cyclic response. Therefore, the
test uses the stress-controlled, cyclic loading module.
The scheme is shown in Table 1.
4
Under the different consolidated pressures, the pore

water pressures increases quickly along the straight
line under the original loading, then the curve bends.
The increase becomes slow. After a long-time, the
increase of pore water pressure becomes smooth.
Finally, the pore water pressure reaches the limit value.
The limit values of the three groups of pore water
pressure are 115 kPa, 160 kPa and 185 kPa, respectively. They are 80% of their effective confining
pressures.
From the above analysis, the increase of pore water
pressure can be divided into three obvious stages: the
sharply increasing stage, the slowly increasing stage
and the smooth stage, as illustrated in Figure 5(c).
1 Sharp increasing stage (AB)
After loading, the excess pore water pressure
increases rapidly in a straight line, and in a short
time (about 1300s, 3250 times of vibration), it
reaches 50% of the limit value. Then the increasing
speed attenuates rapidly with the increasing number
of vibration and at the end of the stage, it reaches a
stable value.
2 Slowly increasing stage (BC)
As a transition stage, the excess pore water pressure comes to a smooth stage from the sharply
increasing stage. The pore water pressure still
increases, but the increasing speed obviously
becomes slow.
3 Smooth stage (CD)
In this stage, the pore water pressure hardly
increases with the increase of the number of vibration. It reaches a stable value called the limit
value, which is about 80% of its effective confining
pressure.
DEVELOPING LAW OF PORE WATER

PRESSURE
Through the test data of GDS, the curve of pore water

pressure with increasing time is obtained, as shown in
Figure 5.
When the characteristics of silty clay are combined

with engineering conditions in Shanghai, Logistic
model of pore water pressure is put forward (the
correlation coefficient reaching over 0.99):
where N is the number of vibration; u0 is the annual

value of hydrostatic pressure; u is the limit value of
excess pore water pressure; N0 and p are regression
Figure 4. Working principle of GDS.

Table 1.
Scheme of undrained dynamic cyclic triaxial test.
Test
number
Sampling
depth(m)
Simulating
depth(m)
Axial
pressure
(kPa)
Confining
pressure
(kPa)
Back
pressure
(kPa)
Dynamic load
amplitude
(kPa)
Circular
stress ratio
CSR
Frequency
(Hz)
D1
D2
D3
8.0 9.0
11.0 12.0
13.0 14.0
8.5
11.5
13.5
155
220
250
130
180
200
80
110
130
0.23
0.7
1.15
0.001
0.003
0.006
2.5
2.5
2.5
The formula of circulars stress ratio: CSR =

2
(Tang, 2003), with =
771
1 + 2 + 3
3
1 + 23
3
(2 = 3 ).
Table 2.
Parameter table of the model.
Number u0
D1
D2
D3
ut
N0
Corresponding
coefficient
78.96 128.22 8125.4 0.43 0.9933

106.44 173.00 6383.2 0.59 0.9979
132.94 191.73 2130.0 0.65 0.9978
pressure under the subway loading is shown in Figure 6. y Axis (unit: kPa) is the response value of pore
water pressure and x axis (unit: s) is time. According to
the calibration coefficients, it is transformed into the
curve of excess pore water pressure with the number
of vibration, as shown in Figure 7.
Compared Figure 7 with Figure 5 (b), the increasing speed of the field monitored pore water pressure
is slower than that of the test, and the increasing
amplitude is also slower. The main cause is the size
effect and the difference between the field test and the
laboratory one. The coefficient c is used to amend the
developing model of pore water pressure in Formula
(1) and the amending model is obtained:
where C is the amending coefficient.

Substituting the field monitored pore water pressures at 8.5 m, 11.5 m and 13.5 m and their number
of vibration into Formula (2), we have the value of c,
0.0350 for D1, 0.2112 for D2 and 0.2975 for D3.
Similarly, the depth of 11.5 m is chosen as an example. The amending parameter is substituted in Formula
(2) and the fitting curve is obtained. The fitting curve
is compared with the monitored curve and the result
is shown in Figure 8. From the figure, the value of
excess pore water pressure in Formula (2) is close to
the monitored value and they are fitting well.
The field monitored data indicate that the excess
pore water pressure is not big when the subway train
is running across. Then it dissipates quickly. The interval of the subway train is 3 5 minutes, when the
excess pore water pressure can almost be dissipated.
That is, basically, the excess pore water pressures produced by the two adjacent trains can not be superposed.
Therefore, the developing law is at the initial stage of
Formula (2) only, that is, the sharply increasing stage.
Figure 5. The increasing pore water pressure with time.
parameters, which are obtained by regression analysis

for test data, as shown in Table 2.
5
FITTING OF FIELD MONITORING DATA
The depth of 11.5 m is chosen as an example. The

changing curve of the field monitored pore water
DEFORMATION OF THE SATURATED

SILTY CLAY
The model can be used to predict the excess pore

water pressure when the subway is running across and
the deformation and the change of stress of soft clay
around the tunnel are analyzed by the effective stress
772
Figure 6. The oscillogram of field pore water pressure at 11.5 m.
Figure 7. The curve of excess pore water pressure and

number of vibration.
Figure 8. The compared curve of field monitoring data and

fitting data.
theory. They offer a theoretical reference to the subway

operation.
The energy of the subway loading is delivered to
the soil by the side wall of the tunnel and the lining.
The sensitivity of pore water is more than that of the
grains skeleton. At the start of the subway loading,
pore water absorbs all the energy of the subway loading, which results in the rapid decline of the effective
principal stress of the saturated silty clay and the elastic discharge of the texture cell of the soil. With the
increasing number of vibration of the running way,
the effective principal stress of the saturated silty clay
reaches a steady value gradually and the texture cell

of soil starts to bear the energy of the subway loading.
The weak connection of the soil cell starts to become
loose and breaks and slight cracks occur, but the soil
cell is still integrated. With accumulating energy, the
crack of the soil cell expands gradually and the big
soil cell splits into small soil cell and crumb. The shear
zone occurs in a place of serious breaking. The small
soil cell and crumb in the shear zone are crushed and
deformation occurs. The accumulation of microcosmic deformation results in the deformation of the axis
of the subway and the ground settlement.
773
The data monitored show that the axial settlement

of the subway Line No. 1 exceeds 20 cm in some tunnel sections, which greatly affects the running of Line
No. 1 and causes some old houses cracking. With the
lapse of time, plastic deformation and large asymmetrical settlement occur on the bottom of the subgrade,
which will affect the subway running and will possibly
result in ground settlement, instability of the building
foundation near the subway, cracking and inclining of
the building, etc. Although no large deformation of
soil occurs ate the side wall of the tunnel in a period
of time, but a perceptible deformation occurs possibly
at the side wall of tunnel with the passing of time of
the subway working.
CONCLUSIONS
This paper uses field monitored data as a parameter for the GDS test. This guarantees the reality of
test conditions. Logistic model of pore water pressure at different depths is put forward. The increase
of pore water pressure can be divided into three obvious stages: the sharply increasing stage, the slowly
increasing stage and the smooth stage. The model is
compared with field monitored data, finding that the
increasing speed of the field monitored pore water
pressure is slower than that of the test and that the
increasing amplitude is also slower. The main cause
is the size effect and the difference between the field
test and the laboratory one. Then, the parameter c is
used to amend the developing model of pore water
pressure and the amending model is obtained. It can
be used to predict the increase of pore water pressure when the subway train is running across. The
effective stress theory can be used to analyse the deformation and the change of stress of silty clay around
the tunnel. The result offers a theoretical reference to
the failure mechanism of the saturated silty clay, the
axial deformation of the subway tunnel and the ground
settlement.
ACKNOWLEDGEMENTS
This work are supported by the research grant
(40372124) from National Natural Science Foundation of China, Shanghai Key Subject (Geotechnical Engineering) Foundation and Shanghai Leading
Academic Discipline Project (Project Number: B308).
REFERENCES
Guan, Q.M., Zhou, S.H. & Wang, B.L. 2004. Variation of
pore pressure and liquefaction of soil in metro.Chinese
Journal of Geotechnical Engineering 26(2): 290292 (in
Chinese).
Guo, Y., Luan, M.T., He, Y. & Xu, C.X. 2005. Effect of variation of principal stress orientation during cyclic loading on
undrained dynamic behavior of saturated loose sands. Chinese Journal of Geotechnical Engineering 27(4): 403409
(in Chinese).
Lee, K.L. & Focht, J.A. 1975. Strength of clay subjected to
cyclic loading. Marine Geotechonolgy 3(2):165168.
Li, L.Y., Cui, J., Jing, L.P. & Du, X.L. 2005. Study on liquefaction of saturated silty soil under cyclic loading. Rock
and Soil Mechanics 26(10): 16631666 (in Chinese).
Meng, Q.S. & Wang, R. et al. 2004. Pore water pressure mode
of oozy silty clay under impact loading. Rock and Soil
Mechanics 25(7): 10171022 (in Chinese).
Pradhan, T.B.S., Tatuoka, F. & Sato, Y. 1989. Stress dilatation
of sand subjected to cyclic loading. Soil and Foundation
29(1):3546.
Shao, L.T., Hong, S. & Zheng, W.F. 2006. Experimental study
on deformation of saturated sand under cyclic pore water
pressure. Chinese Journal of Geotechnical Engineering
28(4): 428431 (in Chinese).
Zeng, C.N., Liu, H.L., Feng, T.G. & Gao,Y.F. 2005. Test study
on pore water pressure mode of saturated silt. Rock and
Soil Mechanics 26(12): 19631966 (in Chinese).
Zeng, Q.J., Zhou, B., Gong, X.N. & Bai, N.F. 2001. Growth
and dissipation of pore water pressure in saturated silty
clay under impact load. Chinese Journal of Rock Mechanics and Engineering 20(1): 11371141 (in Chinese).
Zhou, H.L. & Wang, X.H. 2002. Study on the pore water
pressure of saturated sand in dynamic triaxial test. Journal
of the China railway Society 24(6): 9398 (in Chinese).
774
Analytical solution of longitudinal behaviour of tunnel lining

F.J.M. Hoefsloot
Fugro Ingenieursbureau, Leidschendam, The Netherlands
ABSTRACT: Staged construction of segmented tunnels result in permanent and constant bending moment in
the longitudinal direction. This fact has been confirmed empirically, and analytical solutions for the longitudinal
behaviour of a bored tunnel lining have been presented. This paper summarizes published analytical solutions for
simple loading conditions, and includes corrections for the solutions where necessary. Solutions for additional
loading conditions relevant to TBM tunnel construction are also presented. The analytical solutions have been
built in a powerful Excel spreadsheet for rapid analyses. The results have been validated with staged construction
FEM calculations in PLAXIS 2D. For final validation of the staged construction behaviour, tunnelling data from
the Groene Hart tunnel in the Netherlands were analysed. Segments were instrumented with axial strain gauges,
and the results have been analysed and converted to longitudinal bending moments. It has been demonstrated
that the measured behaviour is reproducible with the analytical model, although selection of input parameters is
complex. The presented model is well suited for quick analyses of TBM back-up train lay out, grouting conditions
and moments from jacking forces with respect to longitudinal lining behaviour.
INTRODUCTION
TBM tunnelling is characterised by phased construction of the segmented lining. Excavation by the tunnel
boring machine is followed by erection of a single
ring of lining segments. The complete structure can
be regarded as a beam on elastic foundation according to classic structural engineering theory. In each
phase of construction, a load free member is introduced, which contributes to the structural system.
In the meanwhile, loading progresses simultaneously
with the TBM. Examples are jacking forces on the first
ring, buoyancy forces within the grouting zone and self
weight of the back-up train.
Previous theoretical analyses of the structural system have shown that the distribution of bending
moment and shear forces is fundamentally different
from a wished-in-place beam on an elastic foundation.
Measurements from the Groene Hart tunnel clearly
confirm this difference.
Bogaards & Bakker (1999) have given analytical

solutions for the section forces of a beam on an
elastic foundation with progressive extension of the
beam under a uniform load. Bakker (2000) published
solutions for items 1 to 3.
Section forces can be derived in several ways.
Bakkers (2000) derivation is explained and the corrected results are given below.
2.2
Uniform load
A staged extension of a uniformly loaded beam can be

solved as the sum of the analytical solutions of a partially loaded beam (Figure 1, left side). The mechanical
2 ANALYTICAL SOLUTIONS
2.1 General procedure
The progressive loading conditions are:
1
2
3
4
uniform loading
shear force at front end of the beam
bending moment at front end of the beam
local uniformly distributed load
Figure 1. Staged construction uniform load.
775
scheme is equivalent to the scheme given on the right

side of Figure 1.
The section forces Mq (x) and Dq (x) (Equation 1 to
4) can easily be summed from the analytical solution
of a beam loaded with a shear force and the bending
moment at the beginning of the beam. These simple
solutions have been given by Hetnyi (1946), Bouma
(1993) and Young et al. (2002).
Figure 2. Staged construction shear force.
2.4 Bending moment at front end of the beam

Progressive extension of the bending moment at the
front end of the beam results in the simple solutions
MM (x) and DM (x) in Equations 9 and 10.
2.5 Local uniformly distributed load

with x = coordinate starting on the left side;
dx = length of beam increment; k = modulus of subgrade reaction; E =Youngs modulus; I = moment of
inertia.
The analytical solution for a progressive local uniform

load can be derived as shown in the mechanical scheme
in Figure 3. The resulting section forces are complex
and are not presented here.
2.3 Shear force at front end of the beam
Figure 2 shows the mechanical scheme of a shear force

at the front end of a stage-constructed beam. As in
Figure 1, the analytical solution of a staged (beam)
extension with shear force at the front end can be
found as the sum of the analytical solution of a partially
loaded beam as explained on the left side of Figure 2.
The mechanical scheme is equivalent to the scheme
given on the right side of Figure 2. The resulting section forces MQ (x) and DQ (x) (Equation 6 to 8) have
been derived from the right side of Figure 2.
The mechanical scheme for the design of a tunnel

lining as a beam on an elastic foundation is given
in Figure 4. An unsupported part of tunnel lining is
present within the TBM (li ) and also between the TBM
and stable grout (lu ). The following loads are present:
SPREADSHEET-MODEL
Bending moment from jack forces (Mjack )

Shear force from jacking forces (Djack )
Shear force from steel brushes (Dbr )
Weight of lining segments (qw )
Uniformly distributed load (q) starting behind the
TBM at distance (l) with a length (lq ).
An Excel-spreadsheet model was built for this

mechanical scheme with the basic solutions given in
paragraph 2. The model is equipped with 6 local,
uniformly distributed loads to account for buoyancy
forces and load configurations of the back-up train,
backfill and permanent structure within the tunnel.
Therefore the model is a powerful tool to calculate
section forces for a wide variety of load conditions
simply by entering the necessary parameters.
776
Figure 5. Angular offset installation new ring.
positioned according to its design alignment, tunnel

rings must be installed with an inclination onto the
previous ring. Figure 5 shows the basic idea of ring
installation.
The spreadsheet-model is equipped with the calculation of angular and vertical deformation based on
this principle. Part of the result is the required installation offset angle relative to the vertical and relative
to the previous ring.
The model has been verified using finite element
modeling with PLAXIS 2D version 8.4. For each of
the basic loading conditions given in paragraph 2.1, a
verification has been performed. In PLAXIS, a structural beam on an elastic soil mass was modeled under
the applicable loading conditions. Phased calculations
have been performed with each phase extending the
beam and soil support, and moving the load one step
ahead. Results form the analytical spreadsheet-model
and finite element model show good to excellent
agreement.
Figure 3. Staged construction local uniform load.
4
Figure 4. Load scheme and subgrade reaction tunnel beam.
Determining the angular and vertical deformation

of the tunnel is not straight-forward. According to
classical structural engineering theory, the angular and
vertical deformation can be found by integration of the
bending moment.
GROENE HART TUNNEL
For the High Speed Railway link between Amsterdam and Brussels, a 7.2 km long TBM tunnel was
recently constructed. The tunnels outer diameter is
14.5 m and the lining thickness is 0.6 m. The Dutch
research committee COB (Center for Underground
Construction) organized an extensive measuring
campaign to study the longitudinal behavior of the
lining. Strain gauges, tilt sensors and two systems to
measure vertical deformation were placed on some of
the tunnel rings.
4.1 Strain gauges
The solution of a stage-constructed tunnel results

in a constant bending moment at great distance from
the TBM. Direct integration of this constant value
results in a linear increase of angular deformation, and
a quadratic increase of vertical displacement. Indeed,
finite element calculations of a stage-constructed tunnel with strain-less extension of the beam show this
type of deformation. Since the final tunnel must be
Strain gauges were used to measure axial and tangential strain during construction. Pairs of gauges were
placed in axial direction, both on the inside and on the
outside of the segments.
After 4 days, the signals reach general equilibrium, with minor fluctuations due to jacking forces.
Complete results from axial gauges for ring number
2117 are shown in Figure 6. The average of the strain
measurements from the inside and the outside of the
segment are given in Figure 7.
777
Figure 8. Axial strain ring 2117, just after having passed

through the TBM, June 3 2003, 1:00 hrs.
Figure 6. Results of all strain gauges ring 2117.
Figure 9. Axial strain ring 2117, at the end of measuring

campaign, June 11 2003, 0:00 hrs.
Figure 7. Average results strain gauges ring 2117.
In Figure 8 typical results of the strain distribution have been plotted against the vertical position,
shortly after having passed through theTBM.The same
has been done at the end of the measuring campaign
(Figure 9). The slope of the trend line through these
measurements changes of sign between the beginning
and end of the measurement program. This axial gradient is determined by the bending moment in the cross
section.
4.2
Bending moment and normal force
Figure 10. Normal force and bending moment from axial

strain measurement ring 2117.
Using the axial strain at tunnel axis and the gradient of

the trend line, the average normal force and bending
moment can be determined with Equations 13 and 14.
with N = average normal force; av = average axial

stress; av = average axial strain; = axial strain gradient; E =Youngs modulus = 38500 MPa; A = section
area = 26.2 m2 ; I = moment of inertia = 634 m4 .
In Figure 10, the derived axial normal force and
bending moment from strain gauge measurements for
the entire logging period are shown. Results of normal

force and bending moment at time of leaving the TBM
have been compared with TBM data of jacking forces.
A reduction factor has been included in the results of
Figure 10. When a reduction factor of 0.9 is applied to
the axial strain, the back-calculated jack forces match
the directly measured jack forces. The reduction factor
takes into account the influence of non-uniform strain
distribution within the segments.
Also the section forces as functions of distance from
the tail end of the TBM are presented in Figure 11.
These lines can only be regarded as normal force and
778
Table 2.
Input parameter; uniformly distributed load.
Load identification
Fig. 4
Load
[kN/m]
Position behind TBM

[m]
qw
q1
q2
q3
q4
q5
629
2064
437.5
70
300
187
6 to 1000
0 to 1000
2 to 26
30 to 1000
52 to 1000
82 to 108
Figure 11. Normal force and bending moment ring 2117.

Table 1.
Input parameter.
Description
Outside diameter
Wall thickness
Youngs modulus
Reduction factor
stiffness
Modulus of subgrade
reaction
Length segments in
tunnel
Unsupported length
Moment jack forces
Shear force jack
Shear force brushes
Identification
Fig. 4
Value unit
14.50 m
0.60 m
3.85E+07 kN/m2
0.650
367,000 kN/m2
li
6m
lu
Mjack
Djack
Dbr
2m
79,000 kNm
0 kN
0 kN
Figure 12. Analytical model and measurement results

ring 2117.
bending moment distribution along the tunnel axis during constant tunnelling. From Figure 11 it is evident
that the bending moment reaches a constant value at
approximately 60 m behind the TBM. This remaining bending moment is a consequence of the staged
construction of tunnelling.
4.3
Back analyses
The analytical model has been applied to calculate the

distribution of the bending moment within the lining
behind theTBM.The input parameters have been given
in Tables 1 and 2. They consist of simple geometric
and material parameters and loading conditions. Two
parameters need further explanation. First, a reduction
factor for the bending stiffness has been applied to
account for the structural behaviour of joints between
segments. Secondly, the external grout load requires
further analyses. Refer to the paper on this subject by
Talmon et al. (2008).
The calculated and measured bending moments
are in Figure 12. There is good agreement between
measurements and calculation results. It is noted that
alternate combinations of input parameters may also

result in fair agreement. However, the analytical model
clearly shows the sensitivity of different loading conditions. Therefore the model is well suited to analysing
loading conditions like position of back-up train,
backfill, jacking forces and grouting conditions.
5
CONCLUSIONS
The mechanical behaviour of a stage-constructed

TBM tunnel is fundamentally different from a wishedin-place tunnel. Analytical solutions for bending
moment, shear force, angular deflection and vertical
displacement have been used to create a spreadsheetmodel. Measurements from the Groene Hart Tunnel
clearly show a residual bending moment far behind the
TBM, a result which is in good agreement with results
of the analytical model. The longitudinal behaviour of
the tunnel as a consequence of the staged construction
needs to be accounted for in tunnel design.
ACKNOWLEDGEMENT
The work described in this paper has been performed
for the research committee F512 of the Centre for
Underground Construction (COB). The author wishes
779
to thank the committee for the opportunity to publish

this work.
REFERENCES
Bakker, K.J. 2000. Soil Retaining Structures. Rotterdam:
Balkema.
Bogaards, P.J. & Bakker, K.J. 1999. Longitudinal bending
in Geomechanics Proc. NUMOG VII: 317321.
Bouma, A.L. 1993. Mechanica van constructies. Delft:

Delftse Uitgevers Maatschappij (in Dutch).
Hetnyi, M. 1946. Beams on Elastic Foundations. Michigan:
The University of Michigan Press.
Talmon, A.M., Bezuijen, A. & Hoefsloot, F.J.M. 2008.
Longitudinal tube bending due to grout pressures.
Shanghai: TC28.
Young, W.C., Budynas, R.G. & Roarke, R.J. 2002. Roarkes
Formulas for Stress and Strain (seventh ed.). New York:
McGraw Hill.
780
Design of tunnel supporting system using geostatistical methods

S. Jeon & C. Hong
School of Civil, Urban & Geosystem Engineering, Seoul National University, Seoul, Korea
K. You
Department of Civil Engineering, University of Suwon, Gyungki-do, Korea
ABSTRACT: Rock mass classification provides a guideline for a tunnel excavation and reinforcement design.
The borehole data and geophysical site investigation results have been popularly used for rock mass classification,
but the locality and limited information from the borehole data and qualitative characteristics of geophysical data
have been problematic. A geostatistical method such as kriging can be an alternative to solve these problems.
This paper describes a design of tunnel supporting system based on geostatistical tools. Korean tunnel supporting
system is typically composed of six different types of combination of shotcrete, rockbolts, and concrete lining
based on rock mass rating (RMR). Ordinary kriging (OK), indicated kriging (IK), and sequential indicator
simulation (SIS) were used to estimate RMR around the tunnel. Kriging methods could estimate RMR with
the best linear unbiased estimator. Using SIS, RMR was presented in the probabilistic term such as mean,
variation, and confidence interval. Reliability of the estimated values was verified by split-sample validation
and compared with the real RMR obtained from the side wall of the tunnel while excavating carried out. Based
on 100 equally probable simulations, RMR could be presented in the form of a probability distribution function
and the uncertainty of estimation could be successfully quantified.
INTRODUCTION
Tunnel excavation and reinforcement design are made

according to the rock mass classification. Engineers
have been using the borehole data of rock mass classification and geophysical site investigation results. Due
to the locality and limited information from the borehole data and qualitative characteristics of geophysical
data, geostatistical method such as kriging should be
considered for the rock mass classification.
Kriging is one of the most widely used interpolation
methods in geostatistics. There has been considerable
researches conducted using this technique (Taboada
et al., 1997; Facchinelli et al., 2001; Marinoni, 2003;
Pardo-Igzquiza and Dowd, 2005). Despite its wide
use, the kriging map flattens out the local details of
the spatial variation with the overestimation of small
values and underestimation of large values. This type
of selective bias is a serious shortcoming because of
the loss of the distribution features of the original data.
Kriging is focused on the estimation of unknown
points by one deterministic value, whereas sequential
simulation is on the stochastic simulation by probabilistic form. Juang et al. (2003) showed the spatial
distribution of soil contamination by the sequential
indicator simulation, and Feng et al. (2006) proposed
an improved sequential indicator simulation.
This paper describes a design of tunnel supporting

system based on geostatistical tools. Typical Korean
tunnel supporting system was composed of six types
of combination of shotcrete, rockbolts, and concrete
lining based on the rock mass rating (RMR). Ordinary
kriging (OK), indicated kriging (IK), and sequential
indicator simulation (SIS) were used to estimate RMR
around the tunnel. For IK, we estimated the threedimensional distribution of RMR with the field data
of borehole logging and geophysical data. And this
result was compared with the results using OK and
SIS. Using SIS, an equally probable simulation was
performed 100 times to quantify the uncertainty of
estimation. The accuracy of estimation was checked
by split-sample validation.
2
ESTIMATION PROCESS
2.1
781
Ordinary kriging
1 Construct a variogram from the scatter point set to

be interpolated.
where h = lag distance; z(x) = value of position x;

and n = number of total data.
2 Define a theoretical variogram. Spherical model

was used in this study.
3 Calculate the weights for each point and estimation
value is the linear combination of weighted known
values.
with a constraint
2
where ab
is the covariance between a and b.
Figure 1. Borehole and seismic velocity data in the research

area.
2.2 Indicator kriging

1 Determine thresholds of borehole data and seismic data. Because borehole data is quantitative and
seismic data is qualitative regarding to RMR, both
borehole and seismic data changed into the indicators between 0 and 1. The thresholds were RMR
20, RMR 40, RMR 60, and RMR 80 for borehole data, and 800 km/s, 1500 km/s, 2400 km/s, and
3600 km/s for seismic velocity.
where vt = indicator value; V (x) = data function.

2 Calculate an indicator of unknown nodes with the
same process of ordinary kriging.
3 Convert an indicator into RMR using cumulative
probability distribution function of estimated four
indicators for four thresholds.
2.3 Sequential indicator simulation
1 Determine the thresholds (1st quartile, medium,
and 3rd quartile) and divide the data into indicators.
The indicator function is given by equation (5).
2 Calculate the experimental variogram and determine the theoretical variogram for each threshold.
3 Select an unsampled node using a random path and
calculate the indicator values at the selected node
by ordinary kriging.
4 Calculate the CDF (Cumulative probability Distribution Function) using three thresholds and
sampling from the CDF.
5 Include the calculated value as conditioning points.
6 Go back to random path selection until all unknown
nodes are calculated.
EXAMPLES AND RESULTS
3.1 Estimation of three-dimensional RMR

Estimation of three-dimensional RMR distribution
was performed in the highway project from Sosa to
Noksan. It was 1700 m in length from STA 3k600 to
STA 5k300, with a depth from 40 to 200 m. The grid
was 86 zones in length (x-direction), 16 zones in width
(y-direction) and 25 zones in height (z-direction). The
dimensions of one element were 20 m in length, 20 m
in width and 10 m in height. RMR estimation was performed by borehole logging data and seismic data.
Borehole location and seismic data were presented in
Figure 1(a) and Figure 1(b), respectively.
The results of ordinary kriging, indicator kriging,
and SIS are shown in Figure 2. Ordinary kriging used
a borehole logging data as input data, and indicator
kriging used both borehole logging data and seismic
survey data. Figure 2(c) shows the first result of 100
SIS results. The RMR distribution around the planned
tunnel is presented in Figure 2(d). Korean tunnel supporting system was composed of six types based on the
RMR. Five grades of RMR are matched with the five
types of support system and sixth support system is for
the portals of a tunnel. Therefore, the most important
issue in the design of tunnel support system can be a
determination of reliable RMR values.
3.2 Reliability analysis of estimated RMR
Split-sample validation was performed to verify the
accuracy of the GA (Genetic Algorithm) simulation.
A subset that was composed of 100 data points from
782
Figure 3. The results of split-sample validation.
Figure 2. Estimation results and RMR distribution abound

the tunnel.
the original borehole-logging data was set aside as test

data; the reminder was training data. And the results of
split-sample validation are presented in Figure 3. The
perfectly estimated result is presented as a straight line
inclined at 45 . The result of split-sample validation
shows dots located around the perfect estimate line.
The dots in the upper and lower parts of the line are
approximately random, and their numbers are almost
identical.
The coefficient of variation was 0.482 and 0.342

for ordinary and indicator kriging, respectively. These
values are very sensitive to local area. With the same
borehole input data, SIS was performed to 100 equally
probable times. Through these analyses of results,
RMR could be presented in the form of a probability
distribution function, and the uncertainty of estimation
could be successfully quantified.
As shown in Figure 3(a) and Figure 3(b), the original input RMR ranged from 15 to 95, whereas the
output RMR ranged from 45 to 80 by both ordinary
and indicator kriging. Distribution features of original geological data were disappeared by kriging in
the process of minimizing the error variation, and this
phenomenon is called as smoothing effect. The coefficient of variation for averaged SIS result was 0.656
as shown in Figure 3(c).
783
of the 100 simulations are presented in Table 1. As

the reliability of the estimation increases, the variance
decreases.
CONCLUSIONS
The objective of this study was to estimate reliable

RMR values and correctly design a tunnel support
system. The results may be summarized as follows:
1 Kriging and sequential indicator simulation have
its special characteristics and sequential indicator
simulation could estimate RMR effectively.
2 Estimation values could be shown in the form
of a probability using the 100 stochastic simulations that were simulated on the condition of
equi-probability. The estimation uncertainty could
be quantified by a variance of RMR.
3 Reliability analysis was performed by split-sample
validation. The differences between true values and
estimation values could check the precision of the
estimation.
Figure 4. Box chart and probability distribution function of

RMR distribution in the planned tunnel area using SIS.
Table 1. Variation of RMR in the planned tunnel area.
Station
Mean
Standard deviation
REFERENCES
3k700
3k940
4k210
4k450
4k690
4k960
5k300
44
62
72
76
75
62
52
16.3
18.7
13.4
12.5
6.5
17.6
21.1
Facchinelli, A. Sacchi, E. & Mallen, L. 2001. Multivariate statistical and GIS-based approach to identify heavy
metal sources in soils. Environmental Pollution 114(3):
313324.
Feng, Y. Tang, S. & Li, Z. 2006. Application of improved
sequential indicator simulation to spatial distribution
of forest type. Forest Ecology and Management 222:
391398.
Juang, K. Chen,Y. & Lee, D. 2003. Using sequential indicator
simulation to assess the uncertainty of delineating heavymetal contaminated soils. Environmental Pollution 127:
229238.
Marinoni, O. 2003. Improving geological models using a
combined ordinaryindicator kriging approach. Engineering Geology 69(12): 3745.
Pardo-Igzquiza, E. & Dowd, P.A. 2005. Multiple indicator
cokriging with application to optimal sampling for environmental monitoring. Computers & Geosciences 31(1):
113.
Taboada, J. Vaamonde, A. Saavedra, A. & Alejano, L. 1997.
Application of geostatistical techniques to exploitation
planning in slate quarries. Engineering Geology 47(3):
269277.
In order to investigate the reliability of the estimation, an equally probable simulation was performed
100 times in order to quantify the estimation uncertainty. The RMR distribution around the planned
tunnel is presented in Figure 4. The left dots and distribution curve represent the RMR realizations, and the
right box chart presents their normal distribution. In
the figure, X, the vertical line, the diamond shape box
chart, and the center dot represent the maximum and
minimum values of 1% and 99% respectively, the box
range from 25% and 75% of the CDF, and the mean
value, respectively. The mean and standard deviation
784
Comparative study of software tools on the effects of surface loads

on tunnels
D.K. Koungelis & C.E. Augarde
School of Engineering, Durham University, Durham, UK
ABSTRACT: In this paper results are present from parametric studies of twin-tunnelling schemes carried out
using various finite element packages. The aim is not to make comparisons with field data but to show the
differences obtained using different FE models, as might be used in a design office. The effects of varying
surface loading on the tunnels themselves are investigated and the changing effects are studied as tunnel layout
is altered.
1
INTRODUCTION
The ongoing development in the worlds urban areas

inevitably leads to the construction of structures in
close proximity to already driven tunnels. Care should
be taken to ensure that construction is carried out
without damaging the tunnels or any other adjacent
or overlying infrastructure. Considerable research has
been undertaken for the case of a single tunnel where
empirical methods for predicting tunnel induced deformations are applicable. For more complex geometries,
however, empirical methods fail to make accurate predictions since they do not account for the soil-tunnelstructure interaction mechanism. The finite element
(FE) method appears to be a solution to this prediction
problem, however many difficulties in its use remain.
The aim of this paper is to validate tools for numerical modelling of tunnelling related interactions in soft
ground.
Plenty of publications exist which study the interaction mechanism of soil, lining and a pre-existing
structure (i.e non-greenfield site) response during tunnelling operations. However the literature on the effect
of surface loading on an existing tunnel is sparse to
the authors knowledge. Most refer to the case of surface loading above pipes, or pile construction and pile
loading, and their effects on tunnels. The main reason
preventing engineers from dealing with the subject of
surface loading is the difficulty they face in accurately
measuring the change of stresses acting on the lining
due to the applied load. The lack of field data results
in performing merely theoretical analysis.
Peck (1969) states three conditions for successful tunnelling. The first refers to safe operation of
tunnelling works. The second requirement is the protection of adjacent structures. The final condition
refers to the tunnels ability to withstand all external
loads which act upon it during its service life. These

loads and their influence on tunnel lining will be considered here. According to ORourke (1984) linings
do not carry the total overburden weight of the overlying ground. The vertical (v ) and horizontal (h )
stresses instead, are re-distributed around the face due
to mobilisation of the soil shear strength and continuity. This effect is often termed arching. The tunnel
therefore has to withstand only the stresses which are
not arched. Mair and Taylor (1997) presented field
data from 12 different tunnel cases driven in London
clay. The lining load is expressed as a percentage
of full overburden weight at tunnel centre line (CL).
The data collection refers to points at one week and
one year after lining installation. These indicate that
the measured lining load even after a year is below
70% of full overburden. In most cases it varies between
40% and 60%.
Moore (1987) described a semi-analytical solution
that makes use of the Boussinesq method and other
closed form solutions to estimate the deformation of a
buried pipe (rather than a tunnel) in an infinite elastic
medium due to surface loading. 2-D FE analysis was
also employed. Provided that realistic elastic ground
properties are selected the semi-analytical method
compares well with the numerical results. This procedure can be used for estimations of hoop forces,
bending moments and ring deformations. Clearly there
are problems however with the assumption of elastic
ground.
This paper investigates the effect of surface loading
on pre-existing tunnels in soft ground assuming plane
strain conditions (2-D FE analysis). The following two
commercial FE packages are used for this purpose,
Strand7 and Plaxis. The purpose of this comparison is
to try and identify the differences in the FE predictions
by using various codes, which might be significant to
785
industrial users of these programs. The reasons for not

using any analytical method (e.g. Boussinesq method)
to estimate the tunnel deformations due to surface
loading were that these methods are only applicable
to elastic medium. They do not take into account the
properties of the medium or the interaction between the
medium and any pre-existing structure in it. According
to Moore (1987) the Boussinesq method fails to take
into account the effect of shear stresses and strains
developing in the overlying strata.
2 ANALYSIS GEOMETRY
In the current plane strain analyses three different tunnel geometric configurations are considered. In the
first case a single tunnel analysis is carried out (ST
case). In the second a twin tunnel analysis is carried
out, where both tunnels are horizontally aligned (TH
case). Finally in the third case twin tunnels are vertically and diagonally aligned (TVD case). A parametric
study was performed for the above three cases varying the position of the tunnel axis (zo ), pillar width
(P), pillar depth (PD ) as well as the position of the
surface loaded area (W ). Figure 1 shows the parameters varied in this study. For the ST case (shown as a
solid circle) zo varies. For the twin tunnel configuration (where the second tunnel is presented as a dotted
circle) P and PD vary. In all cases the loaded area shifts
from W 1 to W 6. Throughout this parametric study the
dimensions of the domain (x, y), tunnel diameter (D),
magnitude (400 kN/m) and area (W ) of the applied
load were constant. Surface load was applied directly
to the surface of the finite elements hence modelling
a flexible footing. No interface elements were used
to model the existence of any type of foundations or
treatment of the ground prior to its loading.
Tunnel diameter (D) was chosen to be 4 m which
is comparable to the diameter of running tunnels for
the Underground in London (Attewell, 1978). The
dimensions of the modelled domain were chosen to
be 70 m long (or 17.5D) in the x direction and 50 m
deep (12.5D) in the y direction. For zo = 15 m and 20 m
the chosen values lie within limits proposed by Potts
et al. (2002). They suggested that for tunnels in clay
the depth of the mesh should be approximately 2D
to 3D below tunnel invert. As for the optimal width
of the domain two factors have to be considered. The
mesh has to be sufficiently wide to ensure minimal displacements along the vertical boundaries. However,
the larger the domain the larger the number of the
degrees of freedom (d.o.f). This immediately affects
the solution in terms of computational time. Hence a
compromise has to be made between these two crucial
factors. It was decided that the above dimensions were
appropriate for this study.
The surface load is constant at 400 kN/m. This magnitude was chosen to resemble the uniform stress from
Figure 1. Geometric parameters of the domain.
a 10-storey building, assuming a stress of 10 kN/m2

per storey for a 4 m wide loaded area. The latter
value over a full building width is probably unrealistic.
However, it was chosen as a worst case value (perhaps including the effect of an accidental concentrated
load) to accentuate the differences in the parametric
study. The value of 10 kN/m2 per storey was chosen
after BS 8002 (British Standards Institution, 1994)
recommendations.
The mesh in Strand7 is created and imported from
Gmsh (a freeware FE mesh generator) since Strand7
can neither produce an unstructured mesh nor can it
be as flexible as Gmsh in the pre-processing stage of
the analyses. Six-noded triangular elements are used
to model the soil, and two-noded beam elements are
used to model the tunnel lining. In Plaxis the autogenerated mesh consists of fifteen-noded triangular
elements. Curved beam elements are used to model the
tunnel lining.The reason for using different type of elements (fifteen-noded triangles rather than six-noded
triangles) compared to Strand7 is to achieve greater
accuracy. In Plaxis the user cannot import a mesh from
another software package. Thus it was entirely created
in Plaxiss pre-processing stage. As a result differences
in the refinement are evident (Fig. 2).
The realistic determination of the initial stress conditions is of great importance in FE modelling in
geotechnics. Several approaches exist for this purpose. The most common of which is the K0 procedure
where stresses prior to any construction are initialised.
This method is only applicable for horizontal ground
surfaces and greenfield sites. This is not the case
in this study. Consequently a different approach is
adopted to simulate initial ground conditions. Tunnel
excavation is not modelled. Instead tunnels with their
permanent lining appear in the mesh as if wished in
place. Then, gravity loading is uniformly applied to the
whole domain (gravity loading method). The resulting
786
Table 1.
Material properties of the soil.
Type of behaviour: Elastic Region, Linear Elastic

Type of behaviour: Yield Surface, Mohr-Coulomb
Parameters
Name
Value
Unit
Youngs modulus
Poissons ratio
Unit weight
Cohesion
Angle of friction
E
v
6.207 103
0.33
20
5
25
kPa
kN/m3
kPa
Table 2.
Material properties of the tunnel lining.
Type of behaviour: Linear Elastic

Parameters
Name Value
Youngs modulus
E
Cross Sectional Area
A
Second Moment of Area I
Unit weight
s
Poissons ratio
vs
Figure 2. Generated mesh for the TVD case when

zo,1 = 15 m for the upper tunnel, zo,2 = 20 m for the lower
tunnel and P = 1D. The upper figure refers to the Strand7
mesh while the lower to the Plaxis mesh.
displacements are then set as the zero datum for the

subsequent steps of the analysis. The stratigraphy is
the same throughout the analysis. This consists of one
clay layer, the characteristics of which are presented in
Table 1. Undrained analyses are performed throughout
using effective stiffness parameters. As for the boundary conditions, in plane strain analysis no horizontal
or vertical movements are permitted along the horizontal boundary at the base of the mesh. On the two
vertical mesh boundaries, only vertical movements are
allowed. The top mesh boundary is free to move.
The beam elements used to model the lining are
assumed to behave in a simple linear elastic way.
Thus two parameters (Youngs modulus E and Poissons ratio v) are required for this model. Table 2
shows the full characteristics of the lining, including the geometrical properties. Soil does not behave
in a linear nor an elastic way. Thus a more realistic
and advanced constitutive model should be adopted.
A simple elasto-plastic constitutive model is therefore
used. For the plastic region Mohr-Coulomb yield criterion with associated flow is used in Strand7 while
non-associated flow is used in Plaxis. A valid criticism here is that this elasto-plastic model is still too
108
0.168
3.95136 104
24
0.3
Unit
kPa
m2 /m
m4 /m
kN/m3
crude for accurate analysis of this problem, which is

true. However the use of this model matches much of
the routine analysis carried out for tunnelling problems in design offices. The purpose of the paper is
not to provide a link to field data but to demonstrate
the different predictions which follow from the use of
different FE models.
In all calculations carried out the analysis procedure began with the tunnels driven and the permanent
lining installed. Displacements from this stage are not
measured. Two load stages are then defined. During
the first gravity load is applied to the mesh. In the following stage the surface load (400 kN/m) is vertically
applied to the pre-defined surface areas (W 1 to W 6).
The displacements due to the first load stage are considered as the zero datum. Thus only those predicted
by the FE analysis due to the second stage (surface
loading) are examined.
3
RESULTS
3.1 Strand7 FE predictions

In this section findings are presented for the case of
surface loading above pre-existing twin tunnels which
are horizontally aligned (TH case). Figure 3 presents
the lining deformations (scaled up) of the first tunnel (Fig. 1) when zo,1 = 15 m and P = 1D due to the
effect of surface loading only (dotted circles). This
is then compared to the original tunnel shape prior
to any loading stage (thick circle). The first obvious
outcome is that the whole tunnel seems to squat. In
other words there is an elongation of the horizontal
787
Figure 3. Deformed shape of the first tunnel (TH case) due

to the surface loading for six different loaded areas. The
tunnel axis is driven at zo,1 = 15 m and P = 1D.
Figure 4. Changes of horizontal and vertical tunnel diameters of the first tunnel (TH case) expressed as a percentage of
the initial tunnel diameter due to the surface loading against
the position of the applied load when zo,1 = 15 m.
diameter with a simultaneous decrease of the vertical.

Further to this obvious vertical translation a secondary
type of movement seems to occur coincidentally. The
deformed lining seems to slightly rotate anti-clockwise
opposing the position of the applied surface load as this
shifts towards W 4. Similar predictions were identified
for the ST and the TVD cases.
Figure 4 shows plots of the change of tunnel diameter (as a percentage of the initial tunnel diameter)
against the relative position of the surface loaded area
for the first tunnel (TH case). The thin lines refer
to the horizontal tunnel diameter (along springlines).
The dotted lines on the other hand refer to the vertical (crown to invert). An increase of the horizontal
diameter with a simultaneous decrease of the vertical
is observed. The maximum increase of the horizontal
diameter as well as the maximum decrease of the vertical (0.8% of the tunnel diameter) occurs when the
surface load is directly applied above for zo,1 = 15 m.
Thus, a clear trend for the magnitudes of these changes
can be identified. These changes appear to fade as
the load is applied further away from the tunnels
centre line.
Figure 5. Plots of the crown settlements due to loading of

the first tunnel in the TH case as a percentage of the ST case
for various surface loading areas and two different tunnel
depths.

the first tunnel (TVD case) as a percentage of the ST case for
various surface loading areas when zo,2 = 20 m.
For the TH case crown settlement predictions from

the first tunnel are plotted (Fig. 5) as a percentage of
the single tunnel case against the position of the surface
loaded area (W ) for two different depths (zo,1 = 15 m
and 20 m) and for three different pillar widths (P = 1D,
2D and 3D). Thin lines refer to the shallow case while
dotted to the deeper. It can be seen that greater interaction occurs for the shallow tunnel case since there
is a difference of 5% to 15% compared to the single
case. For the deeper case results are almost identical
to the single. This implies less or even no interaction
between the two parallel driven tunnels.
For the TVD case crown settlement predictions from
the first tunnel are plotted (Fig. 6) as percentage of
the ST case against the position of the surface loaded
area (W ) for different pillar widths (P = piggy-back,
0 and 1D) when zo,1 = 15 m and zo,2 = 20 m. It can be
observed that a small amount (less than the TH shallow
case) of interaction exist (none to 6%) compared to
the ST results. Greater interaction appears between the
tunnels when they are closely spaced (P = piggy-back
and 0) of approximately 5%. When the lower tunnel is
788
Figure 7. Changes of horizontal and vertical tunnel diameters due to the surface loading expressed as a percentage of
the tunnel diameter against the position of the applied load
when zo,1 = 15 m.
driven further away (P = 1D) less interaction between

the two vertically aligned tunnels occurs (none to 3%).
3.1.1 Comparison of the different cases in Strand7
In this section the FE predictions regarding the first
tunnel (in the TH and the TVD cases) are compared
with the predictions regarding the single tunnel (ST
case) in order to investigate the interaction mechanism
of soil-tunnel-structure in 2-D. The above mentioned
comparison is made in terms of lining deformations.
The FE predictions regarding the first tunnel (TH and
TVD cases) are smaller than those regarding the ST
case. This is a first indication of the existence of interaction. In both of the compared cases the shape of
the tunnels seems to squat while the reduction of the
vertical tunnel diameter equals to the increase of the
horizontal. The maximum lining deformation occurs
when the surface load is applied at W 2. No lining
deformation is predicted when the load is applied at
W 6 (Fig. 4) which indicates that at that distance the
interaction ceases.
In total it seems that interaction occurs within the
region of W 1 to W 4 (P 2D) and then for (P 3D) it
starts to reduce (Fig. 4). Greater interaction is predicted
in the TH case compared to the TVD case. This implies
that the existence of the lower tunnel (TVD case) does
not contribute to the complex interaction mechanism
in the same way as it does the second tunnel in the TH
case (Fig. 1).
3.2 Plaxis FE predictions
Figure 7 is akin to Figure 4. This graph indicates that
the maximum change of D (Roughly 0.8% of the tunnel diameter. These predictions are in agreement with
Strand7) regarding the first tunnel (TH case) occurs
when the load is applied directly above. As the load
shifts further away these changes reduce towards zero
(no change of D). However when the surface load
is applied at its furthest possible distance, the vertical diameter is seen to increase while the horizontal
reduces.
Figure 8. Plots of the crown settlements due to loading of the

first tunnel in theTH case as a percentage of the ST predictions
for various surface loading areas and two different tunnel
depths.

the first tunnel in the TVD case as a percentage of the ST case
for various surface loading areas when zo,2 = 20 m.
Figure 8 shows the interaction between the first

tunnel (TH case) and the surface load compared
to the single tunnel case (ST) for two different
depths (zo,1 = 15 m and 20 m) and three different pillar widths (P = 1D, 2D and 3D) in terms of crown
settlements (similar to Fig. 5). Thin lines refer to the
shallow case while dotted refer to the deeper case. It
can be seen that Plaxis predictions for the TH case are
almost identical to the ST case results (i.e. almost no
interaction between the two parallel tunnels). This picture is different to the Strand7 results especially for the
shallow case.
Figure 9 shows the interaction between the two tunnels and the surface load in the TVD case compared to
the (ST) case in terms of crown settlements for different pillar widths (P = piggy-back, 0 and 1D) when
zo,1 = 15 m and zo,2 = 20 m (akin to Fig. 6). It can
be observed that a small amount of interaction exists
(none to 5%) compared to the ST case predictions.
Results are similar compared to the TH case from W 1
789
flow was used in Strand7 while Plaxis used nonassociated flow. According to Potts and Zdravkovic
(1999) the latter way of modelling real soil
behaviour is more realistic than the first, although
given the loading in this problem it is perhaps not
significant.
to W 3. As the surface load shifts towards W 4 though a

difference in the prediction of interaction is identified.
3.2.1 Comparison of the different cases in Plaxis
FE predictions regarding the first tunnel (in the TH and
the TVD cases) are compared with those in the ST case.
The maximum deformation of the tunnel lining occurs
when the load is applied at W 2 forcing the tunnel to
squat. These deformations reduce as the load shifts
towards W 5. At W 6 though the load seems to produce
an ovalisation of the lining with the vertical tunnel
diameter greater than the horizontal (in contrast to the
previous load cases). In general Plaxis predicts similar
amount of interaction between the two different tunnel
geometric configurations (TH and TVD).
From the above it is clear that differing predictions

can be found from routine use of two FE modelling
packages. Greater detail of this and an extended study
of this problem can be found in Koungelis (2007).
Such differences in predictions may not be important
but without undertaking such studies that fact will be
difficult to assert in most cases.
ACKNOWLEDGEMENTS
DISCUSSION
In the current paper 2-D FE predictions were presented

investigating the effect of surface loading above preexisting tunnels driven in soft ground. Two different
FE packages were used for this purpose to compare
and validate the produced results. Several parametric
studies were carried out varying the excavation depth,
the pillar width, the pillar depth and the position of the
surface loaded area. The general trend between the two
FE packages regarding the lining distortions and the
crown settlements was similar. Both packages predict
that when the surface load is applied within the region
of W 1 to W 4 [i.e. a horizontal distance of P 2D from
the first tunnel (TH case and TVD cases)] the existence
of the interaction mechanism was evident regardless
of the tunnel geometric configuration. Further from
that distance no interaction occurred.
Small differences in the predictions between
Strand7 and Plaxis occurred. Strand7 in particular predicted the existence of stronger interaction mechanism
for theTH case compared to theTVD case. Plaxis on the
other hand predicted a similar amount of interaction
between the two different tunnel geometric configurations and smaller compared to Strand7 predictions.
These differences are attributed to the following three
factors:
Different types of finite elements were used to
model soil in the domain. Six-noded triangles were
used in Strand7. Even though this type of finite
element was available in Plaxis as well, it was
decided that the fifteen-noded triangle should be
used instead for greater accuracy.
Different meshes were generated between the two
FE packages. The reason was that in Plaxis the user
cannot import a mesh as in Strand7.
Finally, even though the same elasto-plastic soil
model with the Mohr-Coulomb failure criterion
was used between the two FE packages, the plastic potential function was different. Associated
This work was funded by a studentship from the UK

Engineering and Physical Sciences Research Council (EPSRC), the School of Engineering at Durham
University and Halcrow Ltd.
REFERENCES
Attewell, P.B. 1978. Ground movements caused by tunnelling
in soil. In: Procs. Of Conf. on large ground movements and
structures, UWIST/July 1977: 812948.
British Standards Institution 1994. BS 8002: Code of Practice
for Earth retaining Structures.
Koungelis, D.K. andAugarde, C.E. 2004. Interaction between
multiple tunnels in soft ground. In: Developments in
Mechanics of Structures and Materials: procs. 18th Australasian Conference, Perth, Australia. Deeks, and Hao,
(eds), London: Taylor and Francis. Vol. 2: 10311036.
Koungelis, D.K. 2007. Tools for numerical modelling of
tunneling interactions. PhD Thesis, Durham University.
Mair, R.J. and Taylor, R.N. 1997. Bored tunnelling in the
urban environment. In: Procs 14th Int. Conf. Soil Mech.
And Found. Engng., Balkema, 1997: 23532385.
Moore, I.D. 1987. Response of buried cylinders to surface
loads. J. Geotech. Engng. 113(7): 758773.
Moore, I.D. and Brachman, R.W. 1994. Three-dimensional
analysis of flexible circular culverts. J. Geotech. Engng.
120(10): 18291844.
ORourke, T.D. 1984. Guidelines for tunnel lining design.
Published by ASCE, Technical committee on tunnel lining
design of the Underground Technology Research Council.
ground. In: 7th Int. Conf. Soil Mech. and Found. Engng,
Mexico, 1969: 225290.
Potts, D.M., Axelsson, K., Grande, L., Schweiger, H. and
Long, M. eds. 2002. Guidelines for the use of the advanced
numerical analysis. London: Thomas Telford.
Potts, D.M. and Zdravkovic, L. 1999. Finite element analysis in geotechnical engineering: Theory. London: Thomas
Telford.
Spasojevic, A.D., Mair, R.J. and Gumbel J.E. 2007. Centrifuge modelling of the effects of soil loading on flexible
sewer liners. Gotechnique 57(4): 331341.
790
Geologic Model Transforming Method (GMTM) for numerical analysis

modeling in geotechnical engineering
X.X. Li & H.H. Zhu
Shanghai, P. R. China
Department of Geotechnical Engineering, School of Civil Engineering, Tongji University, Shanghai, P. R. China
Y.L. Lin
Geotechnical Research Institute, Hohai University, Nanjing, P. R. China
ABSTRACT: During numerical simulation of geotechnical engineering, due to the complex engineering geological condition, the simplified geologic model is usually adopted. However, the accuracy and reliability of
calculating result are directly influenced by characteristic of model. The 3D geologic model can better reflect the
actual geological condition. The paper presents a new and practical modeling method (Geologic Model Transforming Method) of numerical analysis by integrating geologic model and numerical model. This method is
performed with the following procedures: (1) cutting the 3D geologic model according to numerical calculation
region; (2) extracting control data from cutting model to reconstruct the surface model; (3) meshing and forming
the numerical model automatically by the stratum attribute; (4) importing the model into numerical analysis
system. An example is given to illustrate the application of the method. The implementation of the method
results in high efficiency and automaticity of modeling.
INTRODUCTION
With the development of the computer technology,

finite element method (FEM) is widely used in
geotechnical engineering and other research domains.
During the solution of FEM, preprocessing, solving
and the post-processing are included. The preprocessor, which involves collecting data, inputting information, meshing and defining material property, etc.,
is the fussiest process and spends 4050 percent of
the total time in the analysis course of FEM. That is,
about half effort of FEM analysis is expended on the
preprocessor (Liu et al. 2002). Therefore, preprocessing system is one of the core parts of the finite element
analysis and a good preprocessor plays an important
role in the success of the method.
For a practical problem of geotechnical engineering, numerous data is demanded during the modeling
process because of complex geological environment,
structure, construction procedure and so on. If the
fussy data is inputted by manual handling, it is inefficient, and is difficult to check and modify data. Thus,
the simplified geologic model is generally adopted.
However, a poor model may give unreliable results
that are usually misleading and may lead to incorrect
conclusions. Therefore, it is necessary to develop the
research on a simplified good FEM preprocessor, and

the problem has received a lot of interests (Yu et al.
1999, Zhou et al. 2002, Liang et al. 2004).
The information visualization of engineering geology is an important research subject in geosciences. At
present, the geology simulation has the powerful function of 3D geological modeling. 3D strata model can
be established by borehole data, and part spatial analysis can also be operated, such as sectioning and cutting,
etc. There are the certain similarities in modeling and
spatial analysis between geologic model and numerical model. If geologic model can be transformed into
numerical analysis model, its data can also be directly
introduced. Thus, the preprocessor of FEM is largely
simplified (Hou et al. 2002, Xia et al. 2005).
In full consideration of the characteristics of the
geologic model and the numerical model, this paper
presents a new modeling method of numerical analysis called the Geologic Model Transforming Method
(GMTM). This method is suitable for FEM preprocessor in geotechnical engineering and simpler and more
efficient than the traditional operations approach.
The GMTM can be used to form finite element
mesh directly from 3D geologic model. The proposed
technology will give some useful references to the
continuous research on this subject.
791
The remainder of this paper is organized as follows:

in the next section the 3D geologic model is described.
Section 3 introduces the implementation and the algorithm of the new method. Section 4 gives an example
to illustrate the results of this method.
2
2.1
3D GEOLOGIC MODEL
Choosing data model
At present, there are more than twenty data models

presented for 3D geosciences modeling. According
to the form of data structure, these geologic models
presented or discussed could be classified into three
types as facial models, volumetric models and mixed
models. The facial model emphasizes on the surface
representation for terrain, strata interface, outlines of
the constructions, buildings and underground engineeering. The triangulated irregular network (TIN) is
widely applied for the surface modeling (Baker 1989,
Huang et al. 2002). The volumetric model is based
on the spatial partition and the real 3D object construction. The three dimension space can be filled with
regular or irregular 3D volumes, and the representative
volume is irregular tetrahedron network (TEN) (Chen
et al. 1994, Pilouk et al. 1994). The mixed model is the
combination of facial model and volumetric model.
The small storage space of data and its fast visualization are the characteristic of the facial model, and
the information of strata interface and fault could be
represented in the interior of geologic surface model.
The volumetric model is convenient for the description of the attributes of each volume and the storage of
its related spatial location. However, due to the enormous data amount of the solid model, the efficiency
of Boolean operation to strata is lower. Although the
mixed model takes the advantage of facial model for
fast visualization and volumetric model for spatial
analysis in theory, the modeling technology is quite
complicated. By contrast, the paper chooses the 3D
geologic surface model as basic model.
2.2
3D strata surface modeling
The paper is concerned with the 3D geologic surface model based on TIN representation (Zhang et al.
2006). To the geological modeling, the practical implementation just requires:
1. Transaction of borehole data. The information of
drill hole is extracted from database. Then every
borehole is numbered with the dichotomic topologic data structure, and is blocked with the blocking rules of data. Thus, the stratum integrity can be
opened out.
2. Construction of the hollow surface model. The reference TIN is constructed with the surface data of
the borehole. In term of the topological relations of
the formed TIN, each stratum interface is obtained
by mapping the altitude and attribute of under layer

borehole. For the phantom, the modeling can be
established by the sliced interface rules and the
interpolation of the space attribute. So the whole
hollow surface model is finished.
3. Construction of the surrounding surface model.
With the nearest neighbor first algorithm and
Delaunay method to the scattered points in external outline of every interface, the strata surrounding
surface model is constructed. Finally, the 3D strata
surface model is composed of the hollow interface
model and the surrounding model.
3 TRANSFORM GEOLOGIC MODEL INTO
NUMERICAL ANALYSIS MODEL
The finite element mesh needs to coincide with the
consistency principle and the geometrical property
principle. In consideration of the difference in data
structure, the geologic model can not be directly taken
as the numerical model (Wang et al. 2004). To satisfy
the regulation of the finite element mesh, the new modeling method presented in this paper emphasis on the
process of transforming geologic model into numerical
model.
3.1 Flow process
The realized procedures from geologic model to
numerical model are listed in the following:
1. Calculation region cutting (section 3.2). Based on
the 3D strata surface model established, the region
cutting is performed along the partial research
region in conformity to the request of numerical
analysis.
2. Reconstruction of surface model (section 3.3). By
virtue of the intersection points and the strata
control points extracted from the cutting model,
the strata surface model of the calculate region
is reconstructed. Thus, the initial surface mesh is
generated.
3. FEM surface meshing (section 3.4.1). Evaluate
whether the quality of the initial surface mesh is satisfied with the request of the finite element mesh or
not. If the judgment is true, the initial mesh can be
acted as the calculation mesh. Otherwise, the surface mesh needs to be regenerated automatically by
a new approach.
4. FEM solid meshing (section 3.4.2). Based on the
generated surface mesh which satisfied with the
FEM request, the relevant solid mesh is generated
automatically by an efficient approach.
5. Importing into numerical analysis system (section
3.5). The topological data of the FEM mesh model
is imported into the numerical analysis system in
accordance with the requested data format. In the
end, the modeling work is finished gradually.
792
Figure 2. Calculation region cutting.
the intersection lines and the intersection points are

obtained with the intersection calculation of the space
planes each other. By the region cutting, the partial
research region can be arbitrarily acquired on the 3D
geologic model.Thus, it is unnecessary to build the calculation model again and again. At the same time, the
construction process in geotechnical engineering can
be simulated, such as the excavation of the foundation
and the tunnel.
3.3 Reconstruction of surface model
Figure 1. Flowchart of model transformation.
In the Geologic Model Transforming Method, the

key technologies include the cutting, the reconstruction and the meshing. Figure 1 shows a flowchart of the
transforming process. The following sections describe
the key steps in detail.
3.2
Calculation region cutting
According to the engineering geological condition and

the survey data, the 3D strata surface model is established with the geological modeling technology. It has
the advantages of better accuracy for strata attribute
and larger overlying region. To a practical calculation
model, the partial research region just is the region
of numerical model. The conditions which determine
the calculation region usually include removing the
boundary effect and satisfying the calculation accuracy. So it is unnecessary to take the overlying region
of the geologic model as the research region of the
calculation model. To reduce the computing time and
scale of the numerical analysis, the region cutting is
performed in the method. By the means that choose the
calculation region of the numerical analysis, the reasonable cutting boundary could also be determined, as
shown in Figure 2.
To benefit the calculation, the cutting region generally is a rectangular solid. Of course, it can also
be composed with many planes or polygons. The
algorithm of the region cutting is a procedure that
The topological relations are changed when the region

cutting is performed on the original strata surface
model. In order to keep the same topological structure, the surface cutting model is reconstructed by the
intersection lines, the intersection points and the original strata control points. The reconstruction algorithm
includes two sections: the reconstruction of strata
interface and the reconstruction of strata surrounding
surface.
Reconstruction of strata interface:
1. The intersection lines, the intersection points and
the original strata control points in the cutting
region are classified and restored by the strata
attribute.
2. To each stratum, the intersection lines are taken
as the constraint boundary, and then the constraint Delaunay triangularization is applied to the
scattered points in the plane region.
3. The topological structure of the formed mesh is kept
invariability. The altitude of each point is mapped.
Thus, the stratum interface model is obtained again.
4. Repeat step (2) and (3). Each stratum interface
mesh could be reconstructed. Finally, the whole
interface modeling is finished.
The illustration of the reconstruction to the strata
interface is shown in Figure 3.
Reconstruction of strata surrounding surface:
5. The surrounding outline loop of the cutting model
is constructed by the intersection points in the
interface and is oriented counterclock wise.
6. To each cutting plane, the intersection lines of the
plane are taken as the interior constraint lines, and
793
Figure 5. Multiple TIN Regions (MTR) method.
generation need to be judged by the man-machine

interaction way during the process of numerical modeling, such as the preconditions whether the meshing
region is closed or not and whether the total redundant
edge length is zero or not, etc. Thus, the automatic
meshing is often not able to be performed normally.
The reason of the problem does not lie in the algorithms
but in the preconditions. These factors cause the modeling difficulty to be enlarged and the automaticity to
be reduced. Therefore, it is necessary to research how
to improve the whole modeling efficiency entirely.
Figure 3. Strata interface reconstruction.
Figure 4. Surrounding surface strata reconstruction.
then the constraint Delaunay train-gularization is

applied to the intersection points of the cutting
plane.
7. The attribute of the formed mesh in the cutting
planes is determined with the attribute of the
intersection points.
8. Repeat step (6) and (7). The surface mesh of
the closed surrounding cutting planes could be
generated. Finally, the whole surrounding surface
modeling is finished.
The illustration of the reconstruction to the strata
surrounding surface is shown in Figure 4.
The surface model of the cutting region is the
combination of the strata interface model and the surrounding surface model. The cutting model has the
same geometric relation and the topological data structure. Its mesh is also taken as the initial surface mesh
with the strata attribute.
3.4
Finite element meshing
At present, the algorithms of the finite element mesh

generation are rather sophisticated and efficient. Most
relevant researches are focused on the algorithm itself
(George 1991; Shephard et al. 1991; Lau et al. 1996).
However, there are many preconditions of the mesh
3.4.1 Surface meshing

Any single irregular triangle is closed. Such the triangle region is called the Single TIN Region (STR)
in the paper. The set of many STR which do not pass
through each other is called the Multiple TIN Region
(MTR). The surface model based on TIN representation is taken as the MTR model. So the initial surface
mesh of the cutting model also is the MTR model.
By performing circularly the algorithm of the surface
meshing in 3D space plane to each STR, the finite
element surface mesh of MTR model could be generated automatically, as shown in Figure 5. In order
to coincide with topological consistency of the nodes
in shared edge, the region boundaries are discretized
with fixed length partition or integer division way.
The method of the surface meshing is called the MTR
method.
Based on the MTR method, some man-machine
interaction operations can be avoided, such as seeking
manually the closed region and discretizing the boundaries one by one. The MTR method is realized easily
by programming, and has the advantages of better
element quality and high efficiency and automaticity.
3.4.2 Solid meshing
A single closed 3D space composed of the FEM surface mesh is called the Single Space Region (SSR).
The set of many SSR which do not pass through each
other is called the Multiple Space Region (MSR).
The algorithm of the solid meshing needs to be performed in SSR. Generally, there are many different
space regions in a calculation model.Thus, almost each
SSR need to be formed with man-machine interaction.
It is very inconvenient to the numerical modeling.
794
Figure 6. Multiple Space Regions (MSR) method.
So, under the direction of the block thought that

the surface mesh having the same stratum attribute
can constitute the closed space region, a method is
presented to build automatically the SSR of the different stratum. By performing circularly the algorithm of
solid meshing to each SSR, the finite element solid
mesh could be generated automatically, as shown in
Figure 6. In consideration of the dual attributes of the
interface mesh, it needs to be used two times in forming strata space region. The method is called the MSR
method.
The thought of the blocking by strata attribute is
applied in the MSR method. The construction for
the SSR and the geometry inspection can be avoided
before the solid meshing. The efficiency of the whole
modeling is improved thoroughly. It is an innovation
to traditional modeling method.
3.5
Figure 8. Cutting region.
Importing into numerical analysis system
The topological data of the taken FEM surface mesh is

imported into the numerical analysis system in accordance with the requested data format. The numerical
analysis model is formed gradually and the whole
modeling is finished. The finite element model generated with the GMTM is not depended on the numerical
analysis system. Only changing into the demanded
data format, it can be used in the different systems
of the universal FEM software, such as ANSYS and
Marc.
4
Figure 7. 3D geologic surface model.
EXAMPLE
To illustrate the application of the GMTM, an example is presented in this section. The detailed modeling
procedures are the following.
1. The information of boreholes is extracted from
database. The number of boreholes and strata are 16
and 3, respectively. The total number of the strata
control points is 64. The geologic surface model is
constructed by the algorithm of the geologic modeling (see Fig.7). The total number of the TIN is
Figure 9. Topological relations.
144, and the different stratum attribute is indicated

with different colors.
2. The calculation region is determined by importing the information of the cutting surface. The
paper assumes the cutting surface is composed of
four vertical planes. The plane graph of the cutting
region is shown in Figure 8, and the new topological
relations are shown in Figure 9.
3. The cutting surface model is reconstructed according to the strata control points in cutting region. The
strata interface model and the initial surface mesh
are shown in Figure 10.
4. Based on the initial surface mesh, the surface meshing and the solid meshing are automatically performed by the MTR method and the MSR method
respectively. The strata interface FEM mesh and the
solid mesh are shown in Figure 11.
795
Figure 13. Stress distribution: (a)before excavation;

(b) after excavation.
Figure 10. Surface model: (a) strata interface model;

(b) initial surface mesh.
6. The topological data of the formed mesh is

imported into the FEM software Marc. In order to
check the element mesh quality, a simple numerical calculation is implemented. Figure 13 shows the
self-weight stresses of the model before excavation
and after excavation in the action of gravity.
5
CONCLUSIONS
In order to simplify the preprocessor of the numerical

analysis, the paper presents a new modeling method
based on the geologic model. It differs from the traditional modeling method. In our method, the numerical
model is generated by the following procedures: cutting the calculation region, reconstructing the surface
model, finite element automatic meshing, importing
the analysis system. The characteristic of the method
is generalized as:
Figure 11. FEM mesh: (a) strata interface mesh; (b) solid
mesh.
Figure 12. Mesh of the excavation model: (a) whole mesh;

(b) part mesh.
5. The excavation process also can be simulated by the

region cutting. In this example, the cutting region
is taken as the foundation pit. The whole and part
FEM mesh are shown in Figure 12. In the surface
model, the total number of the point and the triangle
element generated are 1140 and 3646 respectively.
In the solid model, the total number of the point
and the tetrahedron element generated are 1821 and
8838, respectively. The total time spent in the whole
modeling process is about 10 minutes.
1. The element mesh is marked by the stratum

attribute, so the model further nears to the real
geological condition. The partial region can be randomly extracted from the 3D strata model, and must
not be established repeatedly.
2. Taking the advantage of the characteristic of the
geologic model, the complex meshing algorithm is
avoided. The programming of the meshing is simple and efficient. The whole modeling is highly
automatic and rapid.
3. The excavation of the foundation and the tunnel
can be conveniently simulated. The boundary of
the excavation must not be set in advance.
4. The finite element mesh can be imported into the
different analysis system to calculate. The model is
applicable and flexible.
5. In a word, the GMTM realizes the dream to establish numerical model directly based on 3D geologic
model. It provides a nice possibility and prospect
for FEM numerical simulation. The transforming
method based on other geologic model is consistent with the method presented in the paper, but the
material algorithm needs to be studied further.
796
REFERENCES
Baker, T.J. 1989. Developments and trends in three dimensional mesh generation. Applied Numerical Mathematics.
Chen, X.Y. & Kozo, I. 1994. Three dimensional modeling
of GIS based on delaunay tetrahedral tessellations. In:
International Archives of Photogrammetry and Remote
Sensing, Munich, Germany, 30: 132139.
George, P.L. 1991. Automatic mesh Generation: Application
to Finite Element Methods. Witey, New York.
Hou, E.K, Wu, L.X. & Li, J.M. 2002. Study on the coupling
of 3D geoscience modeling with numerical simulation.
Journal of China coal Society, 27(4): 388392.
Huang, Y.Y. & Chen, S.Q. 2002. Geologic modeling by
marching cube. Journal of Engineering Graphics, 23(2):
6569.
Lau, T.S. & Lo, S.H. 1996. Finite element mesh generation
over analytical curved surface. Computers & Structures,
59(2): 301309.
Liang, S.W, Yang X.H. & Yang, W.B. 2004. System integration of ANYSIS preprocessor and AutoCAD. Journal of
Huazhong University of Science and Technology (Urban
Science Edition), 21(1): 8184.
Liu, L.M, Liu, H.L. & Zhu, Z.D. 2002. Integration method
of geoengineering finite element analysis system based on
GIS. Chinese Journal of Rock Mechanics and Engineering, 21(Supp.1): 19951998.
Pilouk, M, Klaus, T. & Martien, M. 1994. A tetrahedronbased 3D vector data model for geoinformation.
In: Advanced Geographic Data Modeling, Netherlands Geodetic Commission, Publications on Geodesy,
1994(40): 129140.
Shephard, M.S. & Georges, M.K. 1991. Automatic three
dimensional mesh generation by the finite octree technique. International Journal for Numerical methods in
Engineering, 32: 709749.
Wang, C.X. & Bai, S.W. 2004. Study on integration of 3D
strata information system and FEM. Chinese Journal of
Rock Mechanics and Engineering, 23(21): 36953699.
Xia, Y.H, Bai, S.W. & Ni, C.S. 2005. Study on coupling
of 3D visualization with numerical simulation for powerhouse excavation of a certain hydrojunction. Rock and
Soil Mechanics, 26(6): 969972.
Yu, C, Zhou, X.H. & Zhang, Y.Q. 1999. The integration of
the feature-based modeling and finite element method.
Journal of computer-aided design and computer graphics.
Zhang, F, Zhu, H.H. & Ning, M.X. 2006. Modeling method
of 3D strata suitable for massive data. Chinese Journal of Rock Mechanics and Engineering, 25(Supp.1):
33063310.
Zhou, E.H, Zhu, Y.W. & Wang, T. 2002. Transitionalzone method in hexahedral finite element meshing in
geotechnical engineering. Engineering Journal of Wuhan
University, 35(3): 2429.
797
Review and interpretation of intersection stability in deep underground

based on numerical analysis
T.K. Lu, B.H. Guo, & L.C. Cheng
School of Energy Science and Engineering, Henan Polytechnic University, Jiaozu, Henan, P.R. China
J. Wang
Shenhua Ningxia Coal Group,Yinchuan, Ningxia, P.R. China
ABSTRACT: This paper conducted a preliminary study on the stability of roadway intersection in deep underground conditions on the basis of the underground observations and numerical modeling. Firstly, according to
the intersections used in current coal mines, five different geometrical shapes of two-dimensional intersections
are selected and modeled under deep ground condition. It is noted that the cross-intersection is the most unstable
two dimensional intersection currently used in coal mines. Also, the depth of cover is one of important factors
for this investigation, which clearly indicated that the stability of the intersection is threatened seriously when
the depth of cover reached at 1000 m or more below the surface. In addition, the construction sequence and
direction for the cross-intersection is studied. It is believed that the construction sequence and direction are
sensitive factors for the stability of two dimensional cross-intersection. It is suggested that the formation of the
cross-intersection should be conducted one side by another individually, even this kind of construction sequence
will cause twice stress concentration and redistribution, it is still less influence on the stability of intersection
comparing with other construction methods selected.
INTRODUCTIOIN
From the statistical data, it is noted that the roof failure

mainly occurred around the large span of underground
opening, and the situation is getting worth, when the
complex geometrical structures and deep mining conditions are encountered. This paper is describes the
stability and failure characteristics of intersection on
the basis of literature review and study of the stability and failure behaviour of different geometrical
structure of intersection, which commonly used in
underground of coal mine in China, associated with
various depths of cover using numerical simulation.
It is noted that the geometrical shapes used in China
coal mining industry is much complex, comparing
with major coal mining countries in overseas, such
as Australia and USA. With increasing the depth of
cover, it is susceptible to ground control problems due
to the complex geometries of underground structure.
2
PREVIOUS INVESTIGATION ON
UNDERGROUND INTERSECTION IN COAL
MINES
Stability of underground intersection had been paid

attention previously, even in the shallow condition.
The investigations had been conducted to study

the structural characteristics, stability and failure
bahaviour in the shallow condition by former US
Bureau of Mine (Hanna, 1991).
In 1976, Balachandra studied the stability of intersection using underground monitoring and computer
simulation.
The department of mining engineering from West
Virginia University studied the stability of three-way
intersection with different geometrical parameters,
and it found that the tensile stress is the main reason to
cause the failure of the intersection. Comparatively, the
three-way intersection is more likely having shear failure than the cross intersection. It is concluded that the
tensile and shear failure on the roof strata increased
with the reducing of angle between two roadways
(Peng, 1978).
The department of mining engineering from University of Wollongong also studied the stability of the
T-intersection (Singh, 2001). Based on the numerical
modeling and underground monitoring, it indicated
that the depth of cover, horizontal stress and geometrical parameter are three major factors which
significantly influence the stability of T-intersection.
It also indicated that the stresses induced during intersection formation may result in high incidence of roof
and rib failures.
799
In China, the depth of cover in coal mine is increasing 812 m annually, and there are many collieries
are going to face deep mining situation, with depth
of cover around 10001500 m. With increasing of the
depth of cover, the stress regime and deformation
characteristics of roadway is changed significantly,
it resulted in failure and difficult to ground control.
Also, construction condition is getting worse than
shallow conditions (Jin and Sun, 2001). Under such
situation, the stability of intersection is facing more
serious problem during the deep mining.
The depth of cover is a sensitive factor to the
ground control. For example, the ground problem was
occurred when the depth of cover was changed from
550 m to over 600 m below the surface in Xiezhuang
colliery, Shandong. Previously, the stability of intersection can be maintained by using steel frame when
the depth of cover less than 600 m, but when the depth
of cover reached over 800 m, ground control around
intersection became a major problem, due to: 1) the
soft rock characteristics of surrounding rock mass was
encountered and the mechanical properties was getting worse; 2) the capacity of reinforcing element was
reduced significantly; 3) the plastic zone around intersection increased significantly, 4) roof separation was
formed significantly and it is considered as a major
factor to cause the stability of intersection (Wang,
2001).
The further study on two way intersection indicated
that, in deep underground (700900 m), the angle of
intersection is not a major factor to influence the plastic zone around intersection, but when the angle is
less than 35 degree, the reinforcing area is increased
significantly with decreasing of the angle. It is, therefore, suggested that the angle between two roadways
should not be less than 35 degree for the designation
of intersection (Zhu and Cao, 2005).
On the other hand, many ground control problem
related to intersection are not affected by single factor,
but combination phenomenon, which involved roof
stress, physical properties of mudstone, reinforcement
technique and parameters used are not suitable for such
underground environment (He, et al, 2005).
3
MODELING OF INTERSECTION STABILITY
The modeling was conducted by using FLAC3D with

three dimensional and non-linear simulations. The
numerical model used for the analysis is presented
in Figure 1, hard rock to soft rock, the mechanical
properties are given in Table 1.
3.1
Effect of geometries on stability of intersection
There are large number of roadway intersections are

constructed annually in China coal mines industry with
Figure 1. Model used in the simulation.
various geometries, including three-way intersection,

four-way intersection, two dimensional roadway intersection (Figure 2a,b,c) and three dimensional roadway
intersection (Figure 2d,e) and geological conditions,
e.g. strong rock mass and soft rock mass. With increasing of the mining depth, the stability of these roadway
intersections is susceptible to ground control problems
due to inherently wide roof spans and complicated
intersection geometry used.
The study conducted in this section is to clarify
the effect of the geometrical shapes on the stability of
intersection. To do so, the depth of cover is determined
as 800 m, and the two dimensional intersections with
five different geometrical shapes, plus one normal
single roadway are compared as:
Cross intersection, with 90 angle between two
roadways;
X intersection, similar with cross, but with 45
angle between two roadways;
T intersection;
L intersection;
Y intersection;
Single roadway
The modeling was conducted to evaluate the stability behaviour of the intersection in terms of the roof
and rib deformation and the failure depth into the rock
mass with different shapes, on the basis of mechanical
properties given in Table 1. The deformation and failure of the intersection have been compared with the
normal single roadway, it is indicated that the simpler
the geometry, the more stable the underground structures. The results also shown that the most unstable
two-dimensional intersection is the cross-intersection
(Figures 3 and 4).
800
Table 1.
Mechanical parameters of rock mass.
No
E
(GPa)
12
Modular
of volume
(GPa)
Modular
of share
(GPa)
Density
(kg/m3 )
Cohesion
(Mpa)
Angle
of interfriction (o)
Tensile
Strength
(MPa)
Residual
Cohesion
(MPa)
Residual
angle of interfriction(o)
0.25
4.8
2500
3.7
26
1.4
22
Figure 5. Sequence of intersection construction.
Figure 2. Main intersections used in Chinese coal industry.
Figure 3. Comparison of roof/rib deformation with different

geometrical shapes.
the stability of intersection. As the indicated above,

the cross-intersection is the most unstable two dimensional intersection, thus, the effect of the excavation
sequence on stability of cross-intersection is studied
accordingly.
Figure 5 shows the construction sequences of the
cross-intersection, the sequences are designed as five
different types of a, b, c, d, and e, which represent the
sequences and directions of the construction. According to the modeling results, the minimum vertical
stress and deformation have been found from the excavation sequence Nos. 1 and 2, and both represented
similar construction sequence, that is, the roadways
formed the intersection are developed at different time.
It is indicated that even the stress state around opening are interrupted twice by using these developing
sequences, but the each interruption is considered to be
smaller comparing with other construction sequence
proposed.
The developing direction is also sensitive to the
stability of intersection. If the excavation is toward
the existing opening (Figure 5e and Figures 6a&b),
the significant affect on the stability of intersection
is detached, and if the excavation direction is moving
away from intersection (Figure 5b and Figures 6a&b),
the affect is comparatively small (Figure 6).
3.3 Effect of depth of cover on stability of
intersection
Figure 4. Comparison of roof/rib failure with different

geometrical shapes.
3.2
Effect of excavation sequence on the stability of

intersection
Stresses induced during the intersection formation

may result in high incidence of roof and rib failure
(Singh, 2001), thus the construction sequence (procedure) may be another sensitive factor that influences
The depth of cover is always a sensitive factor, which

should be taken into consideration during the stability
analysis. Based on the previous work, the crossintersection is analyzed with various levels of depth
of cover from 200 m to 1100 m respectively. Figure 7
shows the stability of the intersection various with
depth of cover, and it indicated that the stability is
not affected directly with increasing of the depth of
cover. Before the depth of cover reached a certain
level, such as 1000 m, there is not significant deformation and failure occurred around the intersection,
thus, the stability can be properly maintained. After
801
SUMMARIES AND CONCLUSIONS
Due to inherently wide roof spans used, the stability

of the roadway intersection was always paid attention
by the researchers and engineers. Now, with continual
increasing of the mining depth of cover, more and more
serious stability problems, particularly related to the
roadway intersection, will be faced in Chinese coal
mines.
The study confirmed that the stability of the intersection could be maintained in the shallow condition,
but when the depth of cover reached a certain level,
such as over 1000 m below the surface, the stability of
intersection is under the threat.
The geometry of the intersection is concerned, as
comparatively the geometrical shapes used in China
coal mining industry are relatively complicated, which
results in much more wide roof span than normal
heading roof. Comparing with six different types of
the underground opening, it is noted that, under the
similar conditions, the normal roadway has minimum
displacement and failure than other intersections. It
implies that the large roof span will cause more stability problem than small roof span. Also, comparing
different geometrical shapes of the two-dimensional
intersections, the most unstable one is the crossintersection, which is the most comment intersection
used in coal mines.
The construction sequences also influence the stability of the intersection during the excavation. To
minimize the stress interruption to existing roadway
during the intersection development, particularly for
the cross-intersection, it is suggested that all the additional roadway construction should always move away
from rather than forward to the intersection. In addition, the construction for additional roadway around
the intersection should be conducted one side by
another individually (Figure 5b), even this kind of
construction sequence will cause twice stress concentration and redistribution, but it still is the less
influence on the stability of intersection comparing
other methods of the construction.
The depth of cover is recognized as the most important factor for the stability of underground openings.
According to the modeling results, the stability of the
intersection is not linearly related to the depth of cover,
the significant effect only occurred when the depth
of cover reached to 1000 m under typical surrounding
rock mass conditions selected.
Figure 6. Effect of construction sequences on stability of

intersection.
Figure 7. Effect of depth of cover on stability of intersection.
ACKNOWLEDGMENTS
the depth of cover reached 1000 m, the deformation
and failure is developed significantly, it implies that
reinforcement strategies used for the intersection in
shallow conditions may not be suitably used in the
deep condition.
Authors wish to acknowledge The Innovation Fund

for Outstanding Scholar of Henan Province (Project
No.:0621000400) for providing the financial assistance.
802
REFERENCES
Balachandra, M.B. 1976. The three-dimensional structural
analysis of double-entry and single-entry coal mines.
Volum I: three-dimensional finite element analysis of
crosscut and entry intersection of a double-entry coal
mine, research report, report number(s), PB-80-150345.
Hanna, K. et al. 1991. Coal-mine-entry intersection behavior
study, Report of Investigations.
He, M.C. et al. 2005. Research on Stability and Support
Measures of Crossing Point in Mine Soft Rock Roadway
of Deep Third System, Mine Construction Technology,
Vol.26, No.34, pp. 3235 (in Chinese).
Jin, H.H. & Sun, Q.G. 2001. High stress and soft rock roadway
reinforcement problems and strategy, Coal Technology,
Vol.3, No.3, pp. 3435(in Chinese).
Peng, S.S. 1978. Roof bolting patterns at three-way

entry intersections, research report, report number(s)
NP-23983.
Singh, R.N., Porter, I. & Hematian, J. 2001. Finite element
analysis of three-way roadway junctions in longwall mining, International Journal of Coal Geology, Vol.45, Issues
23, January, Pages 115125.
Wang, P.L. 2001. Discuss of supporting types in roadway
cross area in deep seam, Coal mining, Vol.4, pp. 5759
(in Chinese).
Zhu, Y.S. & Cao, S.H. 2005. Research on Roof Loosed Loop
and Support Reinforced Extent in Roadway Cross Point,
Mine Construction Technology, Vol.26, No.2, pp. 2832
(in Chinese).
803
Analysis of surface settlement due to the construction of a shield

tunnel in soft clay in Shanghai
Z.P. Lu & G.B. Liu
Department of Geotechnical Engineering, Tongji University, P.R.China
ABSTRACT: By comparison of methods of settlement calculations which have been applied into computing
surface settlement due to shield tunneling, the paper analyzed the relativity between Pecks equation and stochastic
medium method, and researched the application problem in practical engineering. Associated with the measured
data of a constructing tunnel, some characters of surface deformation have been analyzed to appraise current level
of construction technology in soft ground in Shanghai. At the basis of established normal ground loss, the quality
of construction can be determined and job practice can be adjusted to meet the requirement of environmental
protection.
INTRODUCTION
Ground movement, which is caused by a shield tunneling in soft clay, will give birth to surface settlement
of different level. When surface settlement arrives
at certain level, the regular service of building and
underground installation around will be influenced.
Therefore, we must find out the regularity of ground
movement, and more exactly predict the value, range of
settlement and maximum gradient of settlement curve.
Besides, in order to adopt some measures reducing
ground movement during construction and design, we
need to analyze all kinds of factors influencing surface
settlement.
Presently, the research about prediction of surface
settlement due to the construction of shield tunnel can
be generalized as follows:
Empirical method;
Theoretical method;
Model experiment method.
This paper presents it most important to predict
the settlement value in allowable error variation by
the most economical and convenient method, which
is used to direct construction process and control
peripheral environmental safety. The final settlement
is mainly caused by ground loss and soil consolidation. And in the period of construction, surface
settlement is mainly caused by soil piling into shield
gap in undrained condition. So based on the comparison of several computing method, the paper firstly
discusses the relationship and applicability of them,

and secondly analyzes the issue of parameter selecting.
In the end, some characters of ground movement and
the calculation of ground loss are presented. To sum
up, less surface settlement indicates less ground disturbance, and the goal is to present reasonable prediction
method and control measures of surface settlement.
2
2.1
REASON AND REGULARITY OF SURFACE

SETTLEMENT
Reason of surface settlement
The reason of surface settlement caused by shield

tunneling can be generalized as follows:
1. Ground loss
Ground movement of working;
Shield setback;
Soil piling into shield gap;
Changing direction of shield driving;
Ground friction and shearing because of shell
movement of shield machine;
Deformation and settlement of tunnel lining.
2. Consolidation of disturbed soil.
2.2
Universal regularity of surface settlement
When shield method is adopted in saturated soft

clay, the universal regularity of ground deformation
along longitudinal axis of shield tunnel is presented in
805
Figure 1. Universal regularity of surface deformation.
Figure 3. Schematic plan of tunnel driving.
some shortages (e.g. considering fewer factor and

lack of theoretical foundation). If value of parameters
exists more difference in common condition, result of
calculation may be influenced obviously.
3.2 Stochastic medium model
Schematic plan of tunnel driving of stochastic medium
model is shown in Figure 3. In this method, after a long
time (t ), the final value of element subsidence
is presented in plane strain state as follows:
Figure 2. Figure of surface settler above tunnel.
Figure 1. Usually ground surface will give birth to little

humping because of uprising and heading of squeezed
soil body in front of shield machine. When shield
machine passes, bilateral soil body will move outward.
When tunnel lining disengages from the end part of
shield machine, ground surface will occur rather large
settlement and the speed rate of settlement is rather
large.
3 ANALYSIS OF SETTLEMENT EQUATION
3.1
Pecks equation
where Se (x) = value of element subsidence in Z level

plane; r(z) = main range of influence due to element
driving in Z level plane; = main angle of influence.
Applying superposition principle, value of crosssection surface settlement (z = 0) due to tunnel driving
may be calculated by
Peck(1969) adopted the following expression to

describe the settlement component due to the construction of a shield tunneling. Pecks equation is expressed
in Figure 2.
where S(x) = value of settlement; Vl = value of ground

loss in unit length of shield tunnel; x = distance to
center line of tunnel; i = spread factor of settler.
Pecks equation expresses the conception that surface settler is approximate normal distribution. And
this estimated equation considers that ground movement is caused by ground loss. Because it is based
on engineering experience, Pecks equation exists
where S(x) = value of cross-section surface settlement; , = area of tunnel driving and tunnel in
unit length; = area of ground loss.
The solution of equation (4) depends on double integral. However, the integrands of these integrals can not
be integrated, so the method of numerical integration
must be applied. Here, value of integral can be gained
by Legendre-Gauss method,
806
COMPARISON OF METHODS OF
SETTLEMENT COMPUTATION
In fact, the gap of tunnel driving (ground loss) must

be defined for several universally adopted method
of surface settlement above. And it corresponds to
separately:
Vl in Pecks equation;
in stochastic medium model;
in Verruijt and Bookers equation
Figure 4. Model of uniform radial displacement.
Assuming that ground loss of tunnel is a infinitesimal element dd.

Calculating by Pecks equation:
where n = number of integral point; Ai , Aj = weight
number of Gauss; 1i ,2j = interpolation points of
integrated variable.
Calculating by stochastic medium method:
3.3 Analysis of computing settlement

Based on theory of ground deformation presented by
Sagaseta (1987), Verruijt & Booker (1996) extend
the theory and present new analytical equation of
ground deformation due to tunnel driving in elastic
half-space, which is also based on equivalent ground
loss model (Lee et al. 1992). Verruijt & Booker
adopted the model of uniform radial displacement
(Fig. 4).
Parameter of equivalent ground loss model is
defined as
where d = external diameter of tunnel; g = parameter

of gap.
Verruijt & Bookers equation is expressed as follows(not considering deformation of tunnel lining in
long-time effects and considering undrained state):
Contrasting equations (10) and (11), if
the two equations are completely identical.

On the one hand, the approximate solution of surface settlement corresponding to different value of x
can be obtained, by the numerical integration of equation (4). Then the curve of surface settlement due to
tunnel driving will be gained by curve fitting of the
result above. On the other hand, regression analysis can
be applied into equation (4) according to the format of
Gauss curve:
The model of ground loss is usually obtained by

assumption, and here surface settlement will be calculated separately by three methods above, adopting
model of uniform radial displacement.
Assuming R = d/2 = 3 m, g = 0.1 m, = 50 , and
surface settlement will be calculated according to two
conditions (z0 /R = 1.3 or 6.7, z0 = distance between
ground surface and center of tunnel). By conversion:
where H = embedded depth of tunnel; = Poissons

ratio; z1 = z H ; z2 = z + H ; r12 = x2 + z12 ; r22 = x2 + z22 ;
m = 1/(1 2).
When z = 0, surface settlement is calculated by
807
By regression analysis, equation (12) is expressed:

If z0 /R = 1.3,
Figure 7. Comparison of computing methods.

If z0 /R = 6.7,
Figure 56 indicate that the computing results exist

less difference between Pecks equation and stochastic medium model if embedded depth of tunnel is high
(e.g. z0 /R > 6). If embedded depth of tunnel is low (e.g.
z0 /R < 2), calculation error is relatively large between
two methods. Besides, the curve of surface settlement
by stochastic medium model corresponds with the
superimposition of Gauss curves of surface settlement
due to series of infinitesimal elements driving. Therefore, it is related with geometric character of working
face, convergence style after driving, not corresponding to Gauss distribution completely in strict meaning
and especially generating relatively large error being
close to maximum settlement. On the other hand,
Pecks equation directly agrees to Gauss distribution.
In a word, stochastic medium theory should be the
theoretical basis of Pecks equation, and Pecks equation is the simplified method of stochastic medium
method, but adopting different methods for parameter

selecting.
Figure 78 indicate that Poissons ratio of Verruijt
and Bookers method locates between 0.1 and 0.2,
adopting relative method of parameters. And its result
is close to Pecks equation when computing maximum
settlement. If embedded depth is relatively high, the
maximum value of settlement is very close.
PRINCIPLE OF APPLICATION OF
COMPUTING METHODS
At present, methods of settlement computation above

have been applied into practical engineering. According to analysis above, it is the key to correctly selecting
parameters for different methods. In other words,
whether method is applied, the separate method of
parameter selecting should be established by experience or theory.
Pecks equation has been applied in the district of
soft soil in Shanghai, because of its characters of simplicity and practicality. And the principle of parameter
808
SZ10-1
SZ10-2
SZ05-3
SZ05
SZ10-3
SZ10
SZ05-4
SZ10-4
SZ15-4
SZ171-4
SZ176-4
SZ181-4
SZ05-5
SZ10-5
SZ15-5
SZ171-5
SZ176-5
SZ181-5
SZ05-6
SZ10-6
SZ15-6
SZ171-6
SZ176-6
SZ181-6
SZ15-1
SZ15-2
SZ15-3
SZ15
Direction of
shield driving
SZ171-1
SZ171-2
SZ171-3
SZ171
SZ176-1
SZ176-2
Construction settelment (mm)
SZ01
SZ05-1
SZ05-2
SZ181-1
SZ181-2
SZ176-3
SZ181-3
SZ176
SZ181 SZ185
Figure 9. The placement of monitoring points.
40
35
30
25
20
15
10
5
0
-8
selecting has been generalized. Spread factor of settler

is expressed as follows (Clough & Schmidt 1981):
-6
-4
-2
10
Maximum head deformation (mm)
Figure 10. The relationship between maximum surface

deformation before shield arriving and maximum surface
settlement after shield heading at the same points.
Stochastic medium theory can consider multiform

factors (e.g. different job practices of tunnel and
sectional styles, etc.), and can calculate multiform
conditions of ground movement (e.g. horizontal displacement of ground surface and surface sloping, etc.).
About parameter selecting, integral domain ( )
and angel of influence () need to be obtained by back
analysis of measured monitoring data.
Verruijt & Bookers method was modified later
(Loganathan & Poulos 1998, Bobet 2001). The result
relatively corresponds with practice, but these methods
still exist some problem (e.g. uncertainty of parameters
of ground loss, curve shape, etc.).
In a word, surface settlement can be predicted by
Pecks equation or analytical equation when solving
problem of parameter selecting according a mass of
statistic of practical engineering. To important and
complicated conditions, stochastic medium method
associated with back analysis is able to gain more
information and more precise resolution.
6.1 Surface settlement caused by ground movement

of working face
Figure 10 generalizes the relationship between maximum surface deformation before shield arriving and
maximum surface settlement after shield heading at
the same points. In fact, they appear to dispersive distribution: surface deformation before shield arriving is
either humping or settlement and corresponding surface settlement after shield heading is either large or
small. Therefore the whole process of surface deformation at certain point is influenced by multiform
factors (e.g. soil layer of tunnel driving, embedded
depth, construction procedure, etc.).
6.2 Final settlement during construction
Vl in equation (1) is calculated by
where Dy = external diameter of tunnel; Vl % = ratio

of ground loss; Vh = ground loss due to ground movement of working face.
Table 12 and Figure 11 indicate:
6 ANALYSIS BASED ON MEASURED

MOINTORING DATA
The paper chooses a constructing tunnel at Shanghai
Metro line 7 as subject investigated. The tunnel adopts
shield method of earth-pressure balance through two
main layers:
Muddy clay belonging to typical soft soil;
Silty clay belonging to hard soil.
The longitudinal direction of tunnel is V in shape,
and the tunnel passes through the crossover facet of
two soil layers. The placement of monitoring points
on ground surface is shown in Figure 9, and the spacing interval of monitoring points is 5 m. The surface
settlement will be separately researched by mainly
considering two factors (ground movement of working face and soil piling into gap of end part of shield
machine).
1. In order to control surface settlement in the district

of complicated surrounding, the method of lowspeed driving and synchronal injecting as soon as
possible can be adopted to gain good result (e.g.
S77,S117).
2. The ratio of ground loss caused by ground humping
before shield arriving in ground loss caused by soil
piling into gap of end part of shield machine is very
little. So ground disturbance before shield arriving
can be controlled at reasonable range at present in
Shanghai.
3. The amount of humping has an effect on the constriction of maximum surface settlement during
construction, but larger humping will give birth to
larger ground disturbance and influence subsequent
value of consolidation settlement.
809
Table 1. Parameters calculation of monitoring points based

on Pecks equation.
Monitoring
points
Z0
m
i
m
Shmax
mm
Vh
mm2
Smax
m
Vl
mm2
SZ47
SZ77
SZ107
SZ117
SZ127
SZ137
SZ157
SZ171
SZ176
14.89
16.80
16.90
16.61
15.85
14.70
13.13
11.95
11.95
6.25
6.88
6.91
6.82
6.57
6.18
5.65
5.24
5.24
1.0
1.2
1.5
2.7
1.8
3.6
1.4
1.7
3.4
0.02
0.02
0.03
0.05
0.03
0.06
0.02
0.02
0.04
17.1
8.50
16.7
5.60
14.5
14.4
13.4
24.3
17.6
0.27
0.15
0.29
0.10
0.24
0.22
0.19
0.32
0.23
Table 2. Controlled maximum settlement and measured

maximum settlement.
Scmax
mm
Smax
mm
SZ47
SZ107
SZ127
SZ137
SZ157
SZ171
SZ176
14.0
12.5
12.0
14.0
16.5
15.3
15.3
17.1
16.7
14.5
14.4
13.4
24.3
17.6
REFERENCES
*Scmax is the maximum allowable value of settlement at

monitoring points.
Distance to the center of tunnel (m)
-20
-15
-10
-5
10
15
20
-5
Surface settelment (mm)
0
5
10
15
20
25
30
CONCLUSIONS
By comparison of methods of settlement calculations

which have been applied into computing surface settlement due to shield tunneling, the paper analyzed
the relativity between Pecks equation and stochastic medium method, and researched the application
problem in practical engineering. Associated with the
measured data of a constructing tunnel, some characters of surface deformation have been analyzed to
appraise current level of construction technology in
soft ground in Shanghai. At the basis of established
normal ground loss, the quality of construction can be
determined and job practice can be adjusted to meet
the requirement of environmental protection.
*Shmax is the maximum value of humping at monitoring

points, and Smax is the maximum value of settlement at
monitoring points.
Monitoring
points
constructed and larger embedded depth, the new

standard of controlling surface settlement is established according to Vl % 1% in Pecks equation.
By analyzing monitoring result, the construction
technology is satisfied with the requirement of controlling surface settlement when the depth of earth
covering is 15 m or so.
SZ10
SZ27
SZ37
SZ47
SZ57
SZ77
SZ87
SZ107
SZ117
SZ127
SZ137
SZ147
SZ157
SZ171
SZ176
35
Figure 11. Final settlement during construction period.
4. Constructing shield-driven tunnel in Shanghai,

30 mm is regarded as controlling standard for maximum settlement (considering that the depth of
earth covering is 4 m). With many metro lines
Bobet, A. 2001. Analytical solutions for shallow tunnels

in saturated ground. Journal of Engineering Mechanics
127(12):12581266.
Clough, G.W. & Schmidt, B. 1981. Design and performance
of excavations and tunnels in soft clay. NewYork: Elsevier
Science Publishing Company.
Lee, K.M. & Rowe R.K. 1992. Subsidence owing to tunneling
I:Estimating the gap parameter. Canadian Geotechnical
Journal 29(6): 929940.
Liu, J.H. & Hou X.Y. 1991. Shield-driven tunnel. Beijing:
China Railway Publishing House.
for tunneling-induced ground movement in clays. Journal of Geotechnical and Geoenvironmental Eengineering
124(9): 846856.
ground. Proceeding of 7th international conference on
soil mechanics and foundation engineering. Mexico City:
State of the Art Report.
Sagaseta, C. 1987. Analysis of undrained soil deformation
due to ground loss. Geotechnique 37(3):301320.
Geotechnique 46(4):753756.
Zhu, Z.L. & Zhang, Q.H. 2001. Stochastic theory for predicting longitudinal settlement in soft-soil tunnel. Rock and
Soil Mechanics 22(1):5659.
810
Urban tunnels in soil: Review of current design practice in Brazil

A. Negro
Bureau de Projetos, So Paulo, Brazil
ABSTRACT: The current design practice of urban tunnels in soil as perceived in Brazil is reviewed and
discussed. The review is based on answers to a questionnaire sent in 2006 to practitioners involved in the design
or in the design supervision of tunnel projects. The results of this investigation are carefully considered to identify
trends and needs for development.
INTRODUCTION
The first assessment of the design practice of urban

tunnels driven in soil performed in Brazil was carried out in 1993 through a survey whose results were
published in the 1st. International Symposium on
Underground Constructions in Soft Ground in New
Delhi (Negro and Leite, 1995).
This time cut-and-cover structures are not included
and the review covers underground openings of any
shape with more than 0.5 m diameter, directional
drillings also not included. The survey is not limited to tunnels built in Brazil and covers any tunnel project conducted with criteria and procedures
presently adopted in Brazil. In all cases tunnels were
designed and built in areas with buildings on the surface and utilities on the subsurface, both liable to
damages induced by ground movements associated to
tunnel excavation.
Multiple choice questions to which one or more
answers could be selected were sent to professionals
involved in the design or in the design supervision
of tunnel projects over the last few years. Considering the multidisciplinary nature of tunnel design, the
questionnaire was sent to practitioners from a variety of expertise and included topics beyond strict
geotechnical context. Accordingly, questions were
sent to geotechnical engineers, geologists, structural
engineers, experts in numerical modelling.
These professionals could opt out questions transcending their specialties. Most questions asked for
answers expressing the practitioner current preference
on each technical aspect of tunnel design, in such a
way that they translated the respondents routine design
practice. The questions were sent to 30 tunnel experts
and 20 replied the survey.
The results presented below are grouped withinrelated topics. One should note that respondents
offered sometimes more than one answer to a question. Therefore, frequencies shown refer to percentage
of total answers offered or to percentage of respondents
as applicable.
2 THE SCENARIO OF THE PRACTICE
To make simpler the analysis of the answers, few questions were formulated to define the scenario of the
design practice to which it refers.
More than two thirds of the answers refer to tunnels
built for trains, metro systems and vehicles in general.
Accordingly, more than four fifths of the answers
(from 26 given) refer to tunnels with more than 4 to 6 m
diameter (almost two thirds with diameters in excess
of 6 m).
The ground condition more frequently encountered, is that of tunnelling with mixed face. Tunnelling
through cohesionless ground is not frequent. On the
other hand, more frequently than not, tunnelling takes
place below water table.
3
LINING DESIGN AND WATER PROOFING
The type of secondary lining more frequently used

over the recent years is the wet mix sprayed concrete (45%), followed in equal proportions by the dry
mix and by the cast in place concrete (23% each).
The answers are consistent with the larger tunnel sizes
focused. Steel liner plates as final lining are not in
use, possibly for being liable to electric or chemical
corrosion in adverse urban scenarios.
The survey revealed that wire meshes are the most
popular type of reinforcement for secondary linings.
It is somewhat surprising to see that lattice girders
and steel sets are still in use, though they are not
usually taken into account in the design of secondary
811
linings. Steel fibers are not much used and it is almost

non existent the old practice of unreinforced concrete secondary linings: it looks like that there is a
general perception that it is necessary to ensure an
adequate ductility rendered by steel reinforcement to
final tunnel linings, assuring that brittle failure of the
lining do not take place under unexpected catastrophic
conditions. This is in line with the requirement of
minimum steel reinforcement accepted by 85% of the
respondents (the remaining do not adopt it).
Forty percent of the respondents explained that
they use only the minimum steel section for bending
whereas 60% adopt the minimum both for bending and
for shrinkage control.
60% of the respondents either never design tunnels
to be waterproof (15%) or they did it only few times
(45%). The most favorite technique to water proof
tunnel linings are impervious geomembranes, system
(39% of the answers) that has been just introduced
in the country. On the other hand this is seconded by
a group of more traditional procedures in Brazil that
aims the reduction of concrete permeability, by special
mix design, special admixtures and controlled curing
of the concrete (summing 32,3% of the answers). This
group combined with the design of concrete with limited crack widths total 45% of the answers, while 5 to
10 years ago they would count for 100% of the answers.
4
Figure 1. Methods of analysis for assessment of tunnel

heading stability.
GEOTECHNICAL DESIGN
All respondents declared they normally perform stability analysis of the tunnel heading. Figure 1 shows
that almost one third of the answers indicate the use
of the safe Lower Bound solutions from the Theory of
Plasticity. Some prefer the use of limit equilibrium
methods or the sole use of Upper Bound solutions
from Plasticity, which are normally unsafe approaches.
Empirical methods (that can either be safe or unsafe)
and finite element analysis with elastic plastic models ranked equally, with almost 14% of the answers
both. It should be noted that within the latter, none
of them were really Numerical Limit Analysis (NLA),
such as that presented by Durand el al (2000). In fact
they refer to 2D FE analysis in which a full ground
stress unloading is applied to check if the 2D opening
withstands it, without any support.
It is worth noting that respondents adopt, in equal
proportion (40%), distinct approaches for stability
analysis of headings of tunnels built with a shield and
without it (mined tunnel, NATM). No formal reason
justifies such practice.
All respondent confirm that they normally predict
the settlements induced by tunnel construction in their
routine practice and that this is done through numerical
analysis (both finite elements and finite differences)
see Figure 2. Comparing these results, with those of
the 1993 survey, a considerable decrease in the use of
Figure 2. Methods for settlements estimates.
empirical and semi-empirical methods is noted (from

60% to 20% now). With respect to the assessment
of induced damages on buildings (Figure 3), no clear
preference to a single method or procedure is noted:
the methods by Burland and Wroth (1974) and by
Boscardin and Cording (1989) were those with more
indications (21% of the answers both). The most used
plane static systems to calculate normal forces and
bending moments in the tunnel lining, accounting the
interaction between it and the ground (see Figure 4) are
numerical solutions with 74% of the answers (48% for
FE and 26% for FD). A considerable increase in the
use of these types of solutions was seen in the country since 1993, when only 43% of the answers favored
them.
It was noted also that in the design of the primary lining, using 2D static systems, 85% of the cases
consider a reduction of the geostatic stresses and the
remaining 15% of answers consider or not this reduction, depending of the specific conditions of the case
being studied. The procedure more frequently used to
estimate the reduction on the ground stresses (Figure 5)
812
Figure 5. Procedures to estimate reduction of ground

stresses.
Figure 3. Methods for assessment of damages induced by

tunnelling on buildings at ground surface.
Figure 6. Account of groundwater loading over a primary

lining of a tunnel below water table.
Figure 4. 2D static systems to assess lining loads.
are those derived numerically (Negro and Eisenstein,

1997, for instance), followed by Terzaghis arching theory, notwithstanding the fact that this corresponds to
large ground stress releases associated to large induced
displacements, which are not compatible to an urban
scenario.
When designing the secondary lining, only 59% of
the practitioners assume a reduction on the ground
stresses, possibly reflecting an old and debatable
believe of long term creep of geological materials,
leading to fully decayed shear strength.
Practitioners were asked how to account for groundwater loads on the sprayed concrete primary lining
of a tunnel driven below water table (see Figure 6).
50% of the answers indicate that no account is taken
of water pressures, on the assumption that the soil is
fully drained (45% of answers) by compulsory dewatering; few (5%) take into account the body forces
resulting from dewatering the ground. The other half
of responses indicate that they take into account the
acting pore water pressures and the effective stresses
onto the lining (either reduced or not).
It is worth noting (see Figure 7) that more than half
of answers offered indicate that practitioners either do
not take into account the effects of pore water pressures in the soil behaviour (4%) or limit the account
to the water pressure onto the lining at most (54%).
Just over 20% perform numerical analysis coupling
soil stresses and pore water pressures using finite elements. Also worth noting (Figure 8) that almost 80%
of the answers indicate the use of elastic-plastic models associated to the Mohr-Coulomb failure criterion.
The problem with this popular option is that considering that we are dealing with an urban tunnel, in which
considerable efforts are spent in reducing the degree
of ground stress relaxation and soil deformation, plastic straining is usually prevented and ground response
813
Figure 7. Account of pore water pressure on soil behaviour.
Figure 9. Choices for geotechnical investigations.
Figure 8. Type of constitutive models used for soil in

numerical analysis.
is reduced to a linear elastic behaviour, an oversimplified portray of a material known to exhibit non linear
behaviour even at small strain range.
Despite the 3D stress redistribution around a tunnel heading, rarely or never (55% of respondents) 3D
numerical analysis are performed in routine design.
When they are performed, they refer to tunnels intersections or start up shafts of tunnels. When 3D analysis
are performed, the most frequently used type of analysis (in 32% of the answers) involve shell elements
for the lining and bars or springs to represent the soil.
3D finite element programs normally used are SAP,
Ansys, Plaxis and FLAC 3D. The more intense use
of 3D analysis already noted in the academy (see the
5th Proceedings of TC28 Symposium in Amsterdam,
2005), resulting from the availability of more powerful
PCs and more efficient numerical codes, will also be
note in industry in the next few years.
GEOTECHNICAL INVESTIGATIONS AND

MONITORING
One of the questions prepared referred to the preference of practitioners with respect to different geotechnical investigations performed in the field and in the
laboratory.
They were asked to indicate, in a scale from 0
(never use) to 3 (always use), their choices regarding each class of investigation. Results of this survey
are shown in Figure 9, from which it can be noted
that: a) standard penetration tests are always requested;
b) cone penetration testing and undisturbed sampling
followed by laboratory special testing are never or
rarely requested; c) disturbed sampling followed by
characterization tests in the laboratory are frequently
used; d) dilatometers, DMT, pressuremeters and plate
bearing tests are rarely used.
With respect to field monitoring, 81% of the
answers confirmed that this is always used in practice
or that it is frequently requested (19%). The answers
indicated also the frequency in using different instruments (Figure 10). Leveling surface settlement points,
building leveling and convergence measurements are
performed in 100% of the cases. Deep settlement
points and lining levelings and the use of piezometers or of water level indicators are observed in 85% to
814
Figure 10. Frequency in using field monitoring.
90% of the cases. Inclinometers, load cells and strain

or stress meters are less frequently used.
6
SATISFACTION WITH THE PRACTICE
Finally, the survey investigated the areas where practitioners believe that there are needs for technical
developments to be introduced in their practice and
areas they are satisfied with the current practice and
new developments are less needed (Figure 11). The
highest levels of satisfaction refer to available construction methods and lining types. The area with least
satisfaction refers to the prediction of pore water pressure and water flow. Practitioners are also unhappy
with the available water proofing techniques and with
the available methods for stability analysis.
7
CONCLUSIONS
The conclusions found refer to the scenario of the practice in Brazil, defined in broad terms by the respondents: large size tunnels, with equivalent diameter
larger than 6 m, designed for rails, metros, highways,
driven under mixed face condition, in cohesive soils,
below water table using sprayed concrete as lining
(NATM).
Steel liner plates are not much in use, and are
being replaced by wet mix sprayed concrete or by
cast in place concrete as secondary lining. This may
be explained by the higher risks of corrosion of the
former in urban environments, but attention should be
Figure 11. Satisfaction with the current practice.
paid to problems related to alkali-silica reaction (ASR)

frequently overlooked in concrete lining design.
It was also noted that the old practice of using
unreinforced concrete secondary lining is declining,
possibly revealing a general perception that the design
should ensure an adequate ductility to the concrete lining, rendered by the steel reinforcement. On the other
hand, it should be argued what is the impact of the steel
reinforcement in the lining safety when subjected to
the high temperatures of a fire.
A considerable increase in the use of water proof
membranes has been observed whereas until recently,
the control of water ingress through the lining was
exerted only by using low permeability concrete.
One wonders if this change may not be followed by
lessening of the concrete quality in the future.
815
It is still high the frequency in the use of limit equilibrium methods and of empirical methods to assess
the stability of the tunnel heading and face, though it
is reckoned that these approaches may provide unsafe
estimates.
Further, an unjustified posture and criterion regarding the assessment of stability of a NATM and a TBM
driven tunnels were noted. For the latter, it looks like
that the tunnel designer delegates to the contractor or
to the machinery manufacturer, the responsibility to
ensure the tunnel face stability and, in general, they
accept it, inasmuch they may not have the technical
abilities to judge what is the best operational procedure
of the equipment at hand, for a certain geotechnical
scene. This subtle omission of the geotechnical engineer is technically and professionally noxious. The
study of cases of frequent instabilities observed in EPB
shields may put some light on this issue.
The use of numerical methods (FE or FD) is
widespread in the local practice for settlements estimates or for lining design. Equally widespread is the
assumption of reduced ground stresses in 2D tunnel
models. However, it is unjustified the persisting practice of using Terzaghis arching theory to estimate that
reduction on geostatic stresses, which is doomed to be
put aside in the practice of urban tunnels.
It is intriguing to observe that practitioners either
do not take into consideration the effect of pore water
changes in the soil behaviour or account only the
water pressure over the lining. FE numerical analysis coupling generation of pore pressure and stresses
are infrequent.
The most used constitutive models in numerical
analysis are those with the elastic plastic behaviour
with Mohr-Coulomb failure criterion. These models
when applied to tunnels built with limited relaxation
of ground stresses, resulting from restrictive construction methods, used to inhibit ground movements and
associated damages in a urban scene, result in linear
elastic soil responses, with inhibited or limited plastic
zones. This is not in line with the reckoned non-linear
behaviour of soils even at small straining: there is a
clear need of using more adequate soil modeling in
the local practice.
The use of three-dimensional modeling in practice
is yet limited, but it is believed that the increased use
of this type of modeling, already noted in the academy,
will be seen in local practice soon.
A poor practice in geotechnical investigation for
tunnel design was noted. It tends to be limited to
soundings with SPT blow counts and simple lab testing in deformed soil samples. This may be so because
we are focusing urban tunnels in environments which
are, in general, well known both in geological and
geotechnical terms, for which sizeable data bank is
available. Perhaps a way to compensate such deficiency is to stimulate the use of in situ testing such
as the pressuremeters, the dilatometers and others.
A favourable situation is seen in the practice, in

terms of field monitoring. Field instrumentation is
always present. The noted deficiency is the lack of
measurements of lining loads. There are a number of
robust and reliable procedures to assess loads in concrete linings, using stress release techniques (see, for
instance, Negro, 1994) that could be used at low costs.
The respondents believe that the area in need for
further technical developments is that of pore water
pressure estimates and ground water flow. Moreover,
they are not satisfied with available techniques for
water proofing tunnels and for assessment of stability analysis of tunnel heading and face. The survey
indicates areas in the practice where the academy may
contribute significantly for technology advancement.
ACKNOWLEDGEMENTS
This investigation has been possible only thanks to the
kind attention of those that replied the questionnaire
sent. The author is grateful to all of them for the efforts
and time spent in replying.
REFERENCES
Bjerrum, L. 1963. Discussion to European Conf. On Soil
Mech. Found. Eng. (Wiesbaden), Vol. II, p. 135.
Boscardin, M.D. & Cording, E.J. 1989. Building response to
excavation induced settlement. Journal of Geotechnical
Burland, J.B. & Wroth, C.P. 1974. Settlement of buildings
and associated damage. Conf. on Settlement of Structures,
London: 611654.
Durand, A.F., Vargas, E.A. & Vaz, L.E. 2000. Some Experiments in 3D Numerical Limit Analysis (NLA). 3rd Intl.
Conf. On Advances of Computer Methods in Geotech and
Geoenvironmental Engineering Geoecology and Computers, Vol. 1: 223227.
Mair, R.J., Taylor, R.N. & Burland, J.B. 1996. Prediction of
Ground Movements and Assessment of Risk of Building Damage due to Bored Tunneling. Proceed. Intl. Symp.
Geotechnical Aspects of Undeground Construction in Soft
Ground, London, UK: 713718.
Namba, M., Ruiz, A.P.T., Queiroz, P.I.B., Negro, A. &
Vasconcellos, C.A. 1999. Assessment of Building Damages Due to Urban Tunnelling. Proc. 11th Panam. Conf.
Soil Mech. And Geotech. Eng., Vol. 2: 549555.
Negro, A. 1994. Soil Tunnels and Their Supports. Special
Conference to the 10th Brazilian Congress of Soil Mech.
Found. Engineering, Iguau Falls, Brazil, ABMS, Proc.
Vol. 5: 3360.
Negro, A. & Eisenstein, Z. 1997. Delayed Lining Activation and Ground Stress Relaxation in Shallow Tunnels.
Proc. of The 4th Intl. Conference on Soil Mechanics and
Foundation Eng, Hamburg, Germany: 23912396.
Negro, A. & Leite. R.L.L. 1995. Design of Underground
Structures in Brazil National Report on Tunnelling and
Braced Wall Excavation in Soft Ground. Proc. Intl. Symposium on Underground Construction in Soft Ground,
New Delhi, India: 4956.
816
A study on loads from complex support system using simple 2D models

Z. Shi, W. Bao, J. Li, W. Guo & J. Zhu
China State Construction Engineering Corporation (SH), Shanghai, P.R. China
ABSTRACT: In deep pit engineering, spatial displacements are readily available, commonly along horizontal
or vertical lines. Its traditionally difficult to get strut loads from complex concrete strut system due to limited
scale of instrumentation and uncertainties in field measurements. This results in problem in predicting wall
deformation where not instrumented. This paper proposes to utilize spatial wall displacements measured to
back-analyze supporting loads on the wall from concrete struts through modeling wall-soil system. The analysis
can be applied to each strut layer to obtain loads between wailing and wall in horizontal plane. The obtained
support loads at different levels are then used as input in a vertical section model, from which deformation
profile of wall can be predicted. The application is verified in project case and shows close correlation to field
measurements.
INTRODUCTION
Reinforced concrete strut system is widely adopted

with diaphragm wall for support of deep foundation pits in soft soil ground. During excavation, the
diaphragm wall is supported by concrete struts from
inside and subject to soil pressure on the other side. It
is typical for deep foundation pits to have a few layers of concrete struts at different excavation depths.
The interaction between diaphragm wall and concrete struts (through wailing), which is one of the
most important factors in analyzing deep foundation
pits, presents a very complicated scenario under loads
from soil.
Because wailing is in continuous direct contact with
diaphragm wall, it is difficult to measure the internal
loads between them. By direct method, the internal
force of struts is commonly measured by embedded
load cells. This measurement is usually insufficient to
estimate support loads on diaphragm wall because:
1 The scale of field measurements of strut force
is limited; and the geometry of strut system is
complex. Therefore, it is difficult to derive reliable support loads on diaphragm wall or through
complex strut system by mechanical analysis with
limited strut force measurements.
2 Field measurement of strut force could be deviated by other factors such as creep, contraction
and thermal stress of concrete material (Xia & Li,
1999). Zhao (1996) reported significant thermal
stress measured in concrete struts.
Therefore, indirect methods are widely used for

analyzing the support loads. Back-analysis from displacement is one of the commonly used. By indirect
method, there have been numerous 2D and 3D studies
on the analysis of such support system (Li & Hu, 1995;
Zhao, et. al., 1996).
In contrary to the difficulty of direct measurement
of support force, spatial displacement information of
diaphragm wall is more readily available and more
reliable by field measurement. Therefore, wall displacement measurements, for example in plan section
at a strut level, can be used to back-analyze loads on
diaphragm wall by a simple wall-soil model.
During excavation, the diaphragm wall deforms
subject to the soil pressure on one side and support
loads on the other side. The soil pressure can be estimated by empirical method. Therefore, for a system
consisting of supports, diaphragm wall and soil, the
measured displacement pattern of diaphragm wall on
the plan can be used to back-analyze supporting loads
on the wall from concrete struts. It is recommended
to carry out the back-analysis in plan section since the
vertical sections normally involve much more complexity associated with construction sequence etc. In
plan of each strut layer, a simple 2D model can easily
represent such case scenario. This can be repeated for
each layer of strut. Therefore, support loads at a section can be back-analyzed from spatial displacement
information from field.
Based on a case history, this paper presents a simple method of estimating support loads on diaphragm
817
wall by wailings. The estimated support loads are then

verified in models for vertical cross section.
2 ANALYSIS OF SUPPORT LOADS IN PLAN
2.1 Assumptions and simplifications
On the plan section of each strut layer, the system of
diaphragm wall and soil resembles a plain strain scenario at the center elevation of wailing. The diaphragm
wall deforms under the support loads and soil pressure after excavation. The sidelines of pit, which in
most cases are straight, can be simplified as an elastic
continuous wall, i.e. a continuous beam on plan. The
deflection pattern of the beam can be easily defined
by displacement measurement along the side and is
taken as target deformation in back-analysis. The two
ends of the sideline can normally be deemed as pinned
ends, as shown in Figure 1.
For simplicity, following assumptions are made in
the analysis:
On plan section, the diaphragm wall is modeled as
elastic continuous beam.
The loads between the wailing and diaphragm wall
are evenly distributed along vertical direction but
vary along horizontal direction for each strut layer.
On plan of each strut layer, the support loads from
wailing are simplified as point loads along the pit
side.
In vertical plan, the support loads of each strut layer
are invariable during construction.
The initial stress level of soil in the model is calculated from its depth following Rankins active earth
pressure formula.
In the plan model, point loads equivalent to the initial stress level prior to excavation are applied onto
the beam (diaphragm wall) so that the model is in initial balance with zero displacement in diaphragm wall.
In back-analysis, the applied point loads are adjusted
from initial values in a controlled manner till that the
diaphragm wall deforms approximately to the target
displacement pattern, which is defined from field measurements. The result point loads can be converted to
internal stress between the diaphragm wall and wailing (Fig. 1 (d)) along the length of sideline. This can be
repeated for each strut layer without major changes to
the model. In this way, the support loads of each strut
layer can be found.
In the model for a vertical cross section, analysis can
be carried out to simulate the construction cases corresponding to plan models. The predicted deformation
pattern of diaphragm wall can checked against field
measurements and the quality of estimated support
loads can be verified.
A history case of a deep foundation pit in soft clay
in Shanghai area is selected for the application of the
Figure 1. Plan section: (a) Plan of strut layout; (b) Line loads
on diaphragm wall from wailing; (c) Converted point loads;
(d) Calculated support stress on diaphragm wall.
method. The program used in the study is the FLAC

by Itasca CG (Itasca, 1997).
2.2 Model conditions and material properties
In the case history, one sideline of the pit with a length
of about 100m is selected for application (Fig. 1 (a)).
The initial stress level is calculated according to
Rankins active pressure with a surcharge of 20 kPa,
as shown in Figure 2 (b). The excavation process is
divided into four steps equaling to the number of strut
layers (Fig. 2 (b)).
818
Figure 3. Model mesh and set up for plan sections.
there are measurements for three monitoring points for

the first strut layer.
According to excavation history, the corresponding
excavation steps for each strut layer are defined as:
up to 2.8 m for the first layer; 6.3 m (2.8 m 8.3 m)
for the second layer; 3.7 m (8.3 m 12.0 m) for the
third layer; and 4.0 m (12.0 m 16.0 m) for the fourth
layer. The initial stress condition for each strut layer is
determined in the same way as above (Fig. 2 (b)).
To reduce noise during back-analysis, only the field
measurements of the center three points (Q10Q12)
are set as target deformation because the other two
monitoring points (Q9 and Q13) could be easily
affected by fixed ends of diaphragm wall. The matching criteria for model prediction is set as within 5%
of field measurement, i.e. the model reaches its target
condition when the model predictions at each selected
points are within 5% difference from field measurements. Under this condition, the point loads on wall
are measured from the model.
The result deformations of back-analysis are shown
in Table 3 with corresponding calculated point loads
of support in Table 4.
Figure 2. Vertical section and simulation conditions.
On each plan section, the loads from wailing

on diaphragm wall are simplified as five point
loads (Fh1 Fh5 ) at the elevation of each strut layer
(Fig. 1 (c)). The target displacements for each support
layer (S1 Si ) are taken at the end of excavation at time tk . The back-analyzed support forces
(Fh1 Fh5 ) represent the site condition at the end
of excavation.
The geometry of the plan model is shown in
Figure 3. The soil layers are listed in Table 1 with material properties. The soil is modeled as elasto-plastic
material with Mohr-Coulomb strength criteria.
The reinforced concrete diaphragm wall reaches
41 m from ground surface, modeled as elastic continuous beam.
2.3 Analysis of support loads at strut layers
There are five monitoring points available for each
layer with field measurement except for the first layer
as listed in Table 2. Due to availability of field data,
3 VERIFICATION OF CALCULATED SUPPORT

LOADS IN VERTICAL SECTION
3.1 Models of vertical section
Two vertical cross sections intersecting the diaphragm
wall at Q10 and Q12 are selected to verify the
calculated support loads.
The vertical section model has a total width of 120 m
with 40 m on the side of the pit as shown in Figure 4.
The height of model is 80 m from the ground surface.
The strut levels and excavation steps are shown in
Figure 2. The model is prescribed with initial lateral
stress conditions as calculated above. The two sides
are fixed only in horizontal direction. The bottom is
fixed in both directions.
Considering the high stiffness and limited height of
wailing, the loads of wailing are assumed evenly distributed along its height at each strut layer. Therefore
in plan model, the support loads can be represented
819
Table 1.
Material properties of soil and diaphragm wall.
No
Soil Layer
Thickness
(m)
Unit Weight
(kN/m3)
Modulus
(MPa)
Poissons
ratio
Cohesion
(kPa)
Friction
( )
12
3
4
51a
51b
51c
6
Top Soil
Mud Silty Clay
Mud Clay
Grey Clay
Grey Silty Clay
Sanded Silty Clay
Silty Clay
Diaphragm Wall
3.5
3
9.9
4.2
7.2
10.4
2.4
1.0
18.3
17.4
16.6
17.5
17.9
17.9
19.9
20.0
11.25
5
2.75
5
12.5
25
36
30000
0.35
0.35
0.4
0.35
0.35
0.3
0.35
0.17
24
15
12
18
17
16
51
15.5
20
11.0
11.5
20.5
23.5
24.5
Table 2.
(mm).
Layer 1
Layer 2
Layer 3
Layer 4
Table 3.
Field measurements of deformation at strut layers
Q9
Q10
Q11
Q12
Q13
15.9
32.3
42.0
0.1
24.3
52.4
62.7
0.3
23.7
49.0
59.9
0.2
19.6
43.5
48.0
13.2
23.3
33.2
Calculated deformations at strut layers (mm).

Q9
Layer 1
Layer 2
Layer 3
Layer 4
Q10
Q11
Q12
0.1
24.3
52.7
63.0
0.3
24.0
50.1
59.1
0.1
19.6
44.1
48.8
Q13
Figure 4. Overview of model for vertical sections.

Table 4.
(kN/m).
Layer 1
Layer 2
Layer 3
Layer 4
Calculated support point loads at strut layers

Table 5.
Q9
Q10
Q11
Q12
Q13
150
89
220
310
106
34
50
150
120
80
220
320
136
39
70
200
108
170
382
540
Converted support loads for vertical model (kN/m).
Section (Q10)
Section (Q12)
by point loads at the center elevation of wailings as

illustrated in Figure 2 (c).
The calculated point loads from plan models are
converted to line loads along horizontal direction at
all strut layers, where the values at Q10 and Q12 can
be obtained (Fig. 1 (d)) for use in vertical model. The
values of point load are listed in Table 5.
In correspondence to the calculated support loads,
the following construction cases are simulated in the
plan model:
Case 1: Initial condition before excavation; installation of diaphragm wall;
Layer 1
Layer 2
Layer 3
Layer 4
318
408
136
156
200
280
300
400
Case 2: Excavate to 2.8 m; apply support load at

first strut layer;
second strut layer;
third strut layer;
fourth strut layer.
3.2 Results of vertical section models
The deformation patterns of diaphragm wall at target
points are very similar as Q12 shown in are shown in
Figure 5.
820
However, the displacement in vertical section shows

notable discrepancies on magnitude, which is partly
a result of simplifications made in the analysis. In
the system of diaphragm wall and soil, the calculated support loads are considerably affected by the
stress conditions in the model. The influence from
stress can be minimized if soil stress measurements
are incorporated in the back analysis.
The discrepancies in prediction of vertical model
could also result from the assumption that the stress
release due to an excavation step is undertaken entirely
by a specific strut layer. Therefore the support loads
can only be calculated for limited cases in the same
number of excavation steps as strut layers.
However, the method produces reasonable estimation of support loads that lead to acceptable prediction
of deformation pattern of diaphragm wall. This would
be helpful in many engineering cases.
ACKNOWLEDGEMENT
Figure 5. Horizontal deformation pattern (mm) of
diaphragm wall along depth (m) in vertical section
through Q12.
The calculated deformations of diaphragm wall are

compared with field measurements in Figure 5. The
maximum calculated displacement is 40.5 mm while
51.1 mm in field measurement on vertical section
through Q12. It is noted that all cases show good correlation of deformation pattern between model calculation and field measurement as shown in Figure 5(b).
4
DISCUSSIONS AND CONCLUSIONS
By using spatial displacement measurements in simple

2D models, the difficulties associated with complex
support system and uncertainties in load measurement
are avoided with reasonable correlation to field data
in vertical sections.
The Author would like to acknowledge great assistance and suggestion in preparation of this paper from
Dr Dongmei Zhang from Tongji University.
REFERENCES
Bolton, M. & Powrie, W. 1988. Behavior of Diaphragm Walls
in Clay Prior to Collapse. Geotechnique, 1988, 38(2):
167189.
Li, Y. & Hu, Z. 1995, Deformation and Loads of Diaphragm
Wall for Deep Foundation Pits. Architectural Construction, Vol 17, 1995(3).
Itasca, C.G. 1993. Fast Language Analysis of Continua.
Itasca Consulting Group Inc.
SIDA, 1997. Shanghai Standard Code for Design of Excavation Engineering (DBJ08 6197).
Xia, C. & Li, Y. 1999. Monitoring Theory and Technology in
Underground Engineering. Tongji University Press.
Zhao et al. 1996. Practice and Study of Support System for
Deep Foundation Pit. Tongji University Press.
821
Ground reaction due to tunnelling below groundwater table

Y.J. Shin
Purdue University, West Lafayette, USA
J.H. Shin
Kon-Kuk University, Seoul, Korea
I.M. Lee
Korea University, Seoul, Korea
ABSTRACT: Tunnelling below the groundwater table influences the hydraulic regime in the surrounding
ground. This will, in turn, cause seepage into the tunnel through the pores and discontinuities and often produces
a long-term interaction between the tunnel and the ground. In this paper an attempt is made to identify the
behavior of surrounding ground due to seepage, and ground reaction curves (GRC) considering seepage forces
are presented for a tunnel under drainage conditions using an analytical method. It is found that the flow of
groundwater has a significant effect on the radial displacement of a tunnel wall. While the effective overburden
pressure is reduced by the arching effect during tunnel excavation, seepage forces still remain. Therefore, the
presence of groundwater induces large radial displacement of the tunnel wall compared to dry conditions.
INTRODUCTION
Tunnelling below the groundwater table affects the

hydraulic equilibrium. This will, in turn, cause seepage
into the tunnel through the pores and discontinuities
in the ground and often produce long-term interaction between the tunnel and the ground. The effect
of seepage on the tunnel is initially reflected in the
ground loading, transmitted through the ground and
then, eventually applied to the tunnel resulting in additional stresses in the linings. It would be appropriate,
therefore, to include the influence of seepage force in
estimating ground behavior due to tunnelling.
The ground behavior due to tunnelling can be indicated theoretically by the ground reaction curve which
shows the increasing trends of radial displacement as
the internal pressure of the tunnel decreases.
In this paper, the theoretical solution of ground reaction curve considering seepage forces due to groundwater flow under steady-state flow was derived. The
studies were performed for a non-supported condition
as well as a supported condition with shotcrete lining.
2
GROUND REACTION CURVE WITH

SEEPAGE FORCES
2.1 Theoretical solutions

It is assumed that a soil-mass behaves as an isotropic,
homogeneous and permeable medium. Also, an
Figure 1. Elasto-plastic model based on Mohr-Coulomb

yield criterion.
elasto-plastic model based on a linear Mohr-Coulomb

yield criterion is adopted in this study, as indicated in
Figure 1:
where, 1 indicates the major principal stress, 3 is the

minor principal stress,
k and a are the Mohr-Coulomb constants, c is the

cohesion, and is the friction angle.
Figure 2 shows a circular opening of radius r0 in
an infinite soil-mass subject to a hydrostatic in situ
stress, 0 .The opening surface is subject to the outward
radial pressure to the tunnel surface, pi . Considering
823
In this state, ir and i are the hydraulic gradient in r

and directions respectively, and w is the unit weight
of groundwater.
If the stress distribution is symmetrical with respect
to the origin (O) in Figure 3, then the stress components are not varied with angular orientation, , and
therefore they are functions of the radial distance r
only. By putting Equation (4) into Equation (2) the
equilibrium equation reduces to the single equation of
equilibrium as follows:
By substituting = 1 and r = 3 in Equation (1)

and by putting Equation (1) into (6), Equation (6) can
be given as follows:
The above partial differential equation can be solved

by using the boundary conditions r = pi at r = r0 .
Then, the radial and circumferential effective stresses
in the plastic region are as follows:
Figure 2. Circular opening in an infinite medium.
Figure 3. Body forces under the groundwater table.
all the stresses on an infinitesimal element abcd of

unit thickness during excavation of a circular tunnel in
Figure 3, equilibrium equation can be expressed as:
If the tunnel is excavated under the groundwater

table, then it acts as a drain in unsupported condition.
The body force is the seepage stress, as illustrated in
Figure 3:
In this equation, pi is all the support pressure developed

by in situ stress and seepage. Subscripts rp and p
indicate the radial and tangential effective stresses in
the plastic region respectively.
In order to estimate the effective stress in the elastic
region, the superposition concept is used. As shown in
Figure 4, the effective stress considering the seepage
force can be assumed as a combination of the solution of equilibrium equation in dry condition and the
effective stress only considering seepage. The Kirsch
solutions are applied to solve the effective stresses in
the elastic region under dry condition (Timoshenko
and Goodier, 1969). And the solution proposed by
Stern (1969) is adopted to obtain effective stresses in
elastic region with consideration of seepage forces.
The following Equation (10) is derived by combining the Kirsh solutions with the Sterns solution in the
elastic region as follows:
824
Figure 4. Concept of superposition in elastic region.
where
Figure 5. The example tunnel for seepage analysis.
Finally, at the interface between the plastic and elastic regions, r = re as shown in Figure 2, the radial stress
calculated in the plastic region must be identical to that
in the elastic region. The following Equation (11) of
the radius of the plastic zone, re can be derived as
follows:
Figure 6. The
H /D = 10).
The radial displacement for a circular tunnel can

be worked out based on the elasto-plastic theory. By
following the same procedure proposed by Sharon
(2003), the expression for the radial displacement in
the plastic region can be obtained as follows:
ground
reaction
curve
(C/D = 10,
The theoretical solution shown above was verified by

performing numerical analysis and comparing the two
analysis results (Shin, 2007).
2.2 Example problem
where,
The radial displacement ur(r=r0 ) at the opening surface

r = r0 is given by Equation (13).
As defined in previous section, the ground reaction

curve, GRC is the relationship between the decreasing
internal pressure, pi and the increasing radial displacement, ur . Given a value of pi , re can be calculated using
Equation (11), then, the ur value using Equation (13).
As shown in Figure 5, the ground reaction curve
is calculated for a sample circular tunnel with diameter of D = 5 m, the cover depth of C = 50 m, under
the ground surface, and the groundwater table of
H = 50 m, above the tunnel crown. As shown in
Figure 6, the ground reaction curve with consideration of seepage force is bigger than the ground reaction
curve in dry condition, which means that there is no
ground water while the cover depth of tunnel, C is
10 times diameter of tunnel, D. This is due to the
fact that even if the effective overburden pressure
825
Figure 7. The distribution of water pressure in fully and

restricted drained cases.
Figure 8. The variation of hydraulic gradient due to lining

set up.
can be decreased by the arching effect during tunnel

excavation, seepage forces still remain.
(13) can be rewritten with the consideration into lining

installation as follows, where rel is the elasto-plastic
interface considering lining installation under groundwater table, assuming the stiffness of the shotcrete is
the same as that of the ground.
2.3 Theoretical solution for lined tunnel when

including consideration into seepage forces
For a real tunnel, the first lining (such as shotcrete),
is installed during tunnelling, and the permeability of
this material is relatively lower than that of ground.
The theoretical solution when considering seepage is
obtained by assuming that the tunnel acts like a drain.
However, the installation of lining makes the groundwater flow change from a fully drained condition to a
partly drained condition.
When the installation of the lining is considered, the
distribution of water pressure in the ground is shown
in Figure 7. The water pressure caused by tunnelling
can be different according to the ratio of permeabilities of the soil and the tunnel lining. In case of the
fully drained condition, the water pressure increases
with depth to a certain depth but it converges smoothly
to zero, uw = 0 at tunnel wall. On the other hand, if
the drained condition is changed to a partially drained
condition due to installation of the lining, then the
water pressure increases with depth like the hydrostatic pressure, but it converges rapidly to zero at the
inside of the tunnel wall. It develops a residual stress
on the surface of lining, therefore eventually induces
a bigger seepage force adjacent to tunnel wall than
that of fully drained condition. For the development of
residual water pressure, a large total head loss is developed within relatively short distance (i.e., thickness of
lining).
When considering the installation of a lining with a
relatively lower permeability than ground, the ground
water flow is changed as shown in Figure 7. This
induces the development of a residual water pressure
on the surface of lining and makes the hydraulic gradient vary. This means that ir in Equation (4) is changed
to ir considering the lining installation. Figure 8 shows
the variation of hydraulic gradient as the result of the
installation of the lining. Therefore, Equation (11) and
The ground reaction curve is estimated through

theoretical solution when considering a lining in which
the permeability is one thousandth of the soil permeability, kl /ks = 0.001, under the groundwater table.
The geometry is same as previous one. Before the lining is installed, the ground reaction curve is the same
as the fully drained curve. Then, after installation of
the lining, a residual water pressure develops on the
surface of lining. It induces an increase of seepage
forces near the tunnel wall, thus the ground reaction
curve increases as shown in Figure 9. Accordingly,
it is known that the permeability and the installation
time of the lining is the most important factor in the
behavior of the ground.
3 THE SIMPLE METHOD OF GROUND
REACTION CURVE
3.1 Estimation of equation of hydraulic gradient
In this study, to evaluate the term related to the seepage in the theoretical solution, the hydraulic gradient
should be estimated in advance under conditions corresponding to the existing geometry. The seepage
analysis is conducted for a sample circular tunnel
826
Figure 9. The ground reaction curve considering lining set

up under groundwater table (C/D = 10, H /D = 10).
Figure 11. The variation of parameter A for water height
and tunnel diameter.
where, ir = the radial component of the hydraulic gradient, z = the depth from the groundwater table to the
tunnel (ex. crown, shoulder, invert), r = the distance
from the center of the tunnel, and D = the diameter
of the tunnel. In general form, Equation (16) can be
expressed as
Figure 10. Generalized hydraulic gradient distribution in

example tunnel.
shown in Figure 5. The radial component of hydraulic

gradient, ir as estimated by the seepage analysis, is
normalized by square of the depth from the groundwater table to the tunnel, z 2 as the sample tunnel shows
in Figure 10, where z is unit of [m], thus unit of ir /z 2
becomes [m2 ]. If the distribution of the hydraulic gradient is expressed as a hyperbolic function, then it can
be easily integrated into the theoretical solution. The
equation of distribution of hydraulic gradient shown
in Figure 10 is as follows:
where, A and B are parameters depending on C/D and

H /C. Shin (2007) performed parametric and sensitivity analysis to assess the two parameters. The results
of his analysis are as follows. Firstly, the parameter B
can be assumed as a constant of one through the sensitivity analysis. In other words, the estimation of the
hydraulic gradient can be reduced to the problem of the
only parameter A. Thirdly, when the groundwater level
is higher than cover depth (H /C > 1), the parameter A
decreases as the groundwater level increases. On the
contrary, when the groundwater level is lower than the
cover depth, the parameter A increases as water level
increases.
Also, when the diameter of tunnel increases, the
parameter A tends to decrease. Thus, if the diameter
of tunnel, groundwater level and cover depth are given
in any geometric condition, the normalized equation
of hydraulic gradient can be estimated by using the
tendency established above. This can be used to estimate the hydraulic gradient necessary to calculate the
ground reaction curve without a seepage analysis in
a preliminary design. Figure 11 is the diagram of the
parameter A including various conditions; it can be
used to estimate the equation of hydraulic gradient in
any geometric condition.
827
3.2
Simplified theoretical solution with

consideration of seepage forces
As mentioned before, the equation of normalized

radial hydraulic gradient can be predicted. The equation of normalized radial hydraulic gradient can be
expressed as follows:
where, C = A zd
Therefore, the elasto-plastic interface and radial displacement can be written as follows, where recs is the
simplified elasto-plastic interface considering seepage
forces.
the arching effect during tunnel excavation, seepage forces still remain. Therefore, the presence of
groundwater induces the large radial displacement
of tunnel wall compared to dry condition.
2 When the shotcrete lining (with a relatively lower
permeability than that of ground) was installed, the
residual water pressure occurs on the surface of the
lining. The seepage forces near tunnel wall increase
and consequently the radial displacement of tunnel increases, too under the assumption that the
shotcrete has the same mechanical properties as the
surrounding soil.
3 The hyperbolic curve (ir = Cr 1 ) that can estimate the distribution of hydraulic gradient due to
tunnelling under groundwater table is suggested
through parametric studies and sensitivity analysis.
Using it, the simplified ground reaction curve can
be achieved without performing seepage analysis.
ACKNOWLEDGEMENTS
This paper was supported by the Korea Institute of
Construction and Transportation Technology Evaluation and Planning under the Ministry of Construction
and Transportation in Korea (Grant C04-01).
REFERENCES
CONCLUSIONS
The groundwater has a significant effect on the behavior of tunnel. This is due to the fact that even if the
effective overburden pressure can be decreased by the
arching effect during tunnel excavation, seepage forces
still remain. The results obtained from this study can
be summarized as follows:
1 The flow of groundwater has a significant effect
on the radial displacement of a tunnel wall. While
the effective overburden pressure is reduced by
Sharan, S.K. 2003. Elastic-brittle-plastic analysis of circular

openings in Hoek-Brown media, International Journal of
Rock Mechanics and Mining Sciences 40: 817824.
Shin,Y.J. 2007. Elasto-plastic ground response of underwater
tunnels considering seepage forces, Ph.D. Thesis, Korea
University, Seoul
Stern, M. 1965. Rotationally symmetric plane stress distribution, Zeitschrift fur Angewandte Mathematik und
Mechanik 45(No. 6): 446447.
Timoshenko, S.P. & Goodier, J.N. 1969. Theory of elasticity,
McGraw-Hill, New York.
828
Basal stability of braced excavations in K0 -consolidated soft clay by

upper bound method
X.Y. Song & M.S. Huang
ABSTRACT: Design of braced excavations in soft clays is usually controlled by the short-term undrained
stability. The paper takes into consideration the principal stress axial rotation induced by excavation and undrained
anisotropic soil strength.Assuming the Prandtl soil slip failure modes and considering the anisotropic soil strength
recommended by Casagrande & Carillo and the non-homogeneous feature of soft soils, a method for evaluating
the basal stability is proposed based on the upper bound analysis. Results obtained from the proposed method
indicate that the basal stability is significantly influenced by the anisotropy ratio of soil, as well as the plane
geometry of the excavation and the thickness of soft soil layer between excavation base and hard stratum. For
field case studies, the proposed method is testified by field observations as well as finite element methods.
INTRODUCTION
For deep excavations in soft clay, design of the lateral

earth support system is often controlled by stability
requirements. In current practice, basically there are
three methods available for performing stability calculations of braced excavations: (1) limit equilibrium
methods; (2) displacement-based elastoplastic finiteelement methods; (3) upper and lower bound limit
analysis. The limit equilibrium methods are widely
used in design practice and include separate calculations of basal stability (based on failure mechanisms
proposed by Terzaghi 1948; Bjerrum and Eide 1956)
or overall slope stability (using circular or noncircular
arc mechanisms) based on well established methods
(Morgenstern and Price 1965; Bishop 1966; Spencer
1967). It is often difficult to assess the accuracy
of these solutions due to ad hoc assumptions: (1)
in selecting the shape of the failure surface; (2) in
the search procedures used to locate the critical surface; and (3) in the approximations used to solve the
equilibrium calculations (Ukritchon et al., 2003). Further complications arise in analyzing soil structure
interactions for embedded supported walls, tieback
anchors, etc.
Displacement-based elastoplastic finite-element
methods provides a comprehensive framework that can
evaluate multiple facets of excavation performance
ranging from the design of the wall and support system, to the prediction of ground movements, and the
effects of construction activities such as dewatering,
ground improvement, etc. They are indispensable for

predicting the distribution of ground movement caused
by excavations, and for simulating process where there
is partial drainage within the soil. Excavation stability
is usually assessed by factoring the strength parameters
of the soil.
A powerful method of calculating stability is based
on the upper and lower bound theory. Chang (2000)
analyzed the safety factor of basal stability based on
the Prandtl slide failure mode using upper bound theory. Ukritchon et al.(2003) formulated the numerical
limit analysis by assuming the anisotropic yield criterion proposed by Davis and Christian (1971). Zou
(2004) assumed a kind of complex velocity field and
analyzed the basal stability without considering the
anisotropic feature of K0 consolidated clay.
However, the undrained shear strength of natural
K0 -consolidated clay is anisotropic and has a close
relation with the vertical effective stress. The classical
rotation of principal stress direction of clays induced
by excavation is illustrated in Figure 1 (Clough and
Hansen, 1981) and it is well-known that the undrianed shear strength of clay varies with the angle of the
major principal stress reorientation during the loading.
Some researchers (Jiang et al., 1997; Su et al., 1998;
Ukritchon et al., 2003) pointed out that without considering the anisotropic and non-homogeneous feature of
clays the result of basal stability is not realistic and the
safety factor may be much smaller than the real one.
In this paper, an upper bound analysis method for
evaluating the basal stability of deep excavations in
829
Table 1.
results.
Comparison between authors and measured
Test
method
Undrained shear
strength expression
Test value
(Ladd, 1973)
Theoretical
value
K0 UC
K0 UE
eq. (7)
eq. (8)
0.330
0.155
0.367
0.140
The undrained limit shear strength is the intersection of undrained stress path and critical state stress
line. If let = M , the undrained limit shear strength
obtained from compressive test can be expressed as
Figure 1. Typical rotation of principal stress.
soft clay is introduced by taking into consideration of

anisotropic feature of K0 consolidated soft clays in the
world, especially in Shanghai, as well as the undrained
shear strength variation with vertical effective consolidation stress. Practical application of the proposed
method is demonstrated through case studies using
data published in literature.
2
2.1
Similarly, if let = M , the undrained limit shear

strength obtained from triaxial extensive test can be
expressed as
where pc = mean effective stress; = the second

invariant of anisotropic tensor: ij . In the analysis, the
initial value of ij is determined by the initial consolidation state of the soil. For the K0 -consolidated clay,
pc can be expressed as
UNDRAINED ANISOTROPIC SHEAR

STRENGTH OF K0 -CONSOLIDATED
SOFT CLAYS
Undrained shear strength obtained from
triaxial tests
Wei and Huang (2006) presented the detailed formulation of a constitutive model for soft clays which can
consider the anisotropic feature of clays based on the
bounding surface plasticity. In the case of the shape
parameter R = 2, the volumetric strain rate can be formulated from the yield function and hardening rule
which can be expressed as follows.
where = q/p = stress ratio; e0 = initial void ratio; ,

= the slopes of virgin consolidation line and swelling
line respectively in the e ln p space; M = critical
stress ratio; and = slope of yield surface in p q
space.
Under the triaxial undrained condition, i.e. v = 0,
equation (1) can be rewritten as
By integrating both sides of equation (2), the

undrained stress path can be expressed as

where v0
= vertical effective consolidation stress.
Substituting equation (6) and Su = qult /2 to equations (4) and (5) respectively leads to the following
expressions as
and
where Suv = undrained shear strength obtained from

K0 UE triaxial test; and Suh = undrained shear strength
obtained from K0 UC triaxial test.
Table 1 shows the comparison between the aforementioned formulas and experimental data from
Ladd(1973), which testifies that the current expression of undrained strength for soft clays is suitable.
According to Jiang et al. (1997), 1/ = 0.76. Other
input parameters are as follows:
K0 = 0.5, = 33 , = 3(1 K0 )/(1 + 2K0 ) = 0.75,
M = 6sin /(3 sin ) = 1.331.
830
direction caused by excavation can then be described

as follows:
Taking advantage of the definition of safety factor

by strength reduction finite element method, we can
define the basal stability factor as
where Su (z) = real undrained shear strength of clays;

and Su (z)critical = critical undrained shear strength.
Figure 2. Definition of geometric parameters.
2.2
Undrained shear strength considering the

rotation of principal stress direction caused by
excavations
In order to consider the rotation of principal stress

direction caused by excavation in clays, the classical
anisotropic strength formula for soft clays recommended by Casagrande & Carillo (1944) is adopted
here. In the vertical plane, the undrained shear strength
under any principal stress direction is expressed as
where Sui = undrained shear strength of clays when the

angle between principal stress direction and vertical
direction is i. Suh and Suv can be obtained from general
undrained triaxial tests.
According to the geometric relationship in Figure 2,
i can be express as
where = angle between the direction of soil failure

surface and vertical direction; = angle between the
direction of soil failure surface and the principal stress
direction.
According to the test result from Lo (1965) is
a constant which does not vary with the rotation of
the principal stress direction. For an undrained analysis, = /4 is preferred. In addition, the anisotropic
strength ratio is defined as k = Suh /Suv . For an
isotropic clay Suv = Suh , k = 1. Substituting the expressions of k and results in the following expression,
By substituting equation (4) to equation (11), the

anisotropic strength considering both the vertical
effective stress and the rotation of principal stress
UPPER BOUND ANALYSIS OF BASAL

STABILITY OF EXCAVATION
CONSIDERING UNDRAINED
ANISOTROPIC SHEAR STRENGTH
OF CLAYS
3.1 Soil slide failure mechanism

Chen (1975) strengthened that the assumed soil slide
failure mode is crucially important to the final result
of the limit analysis. Chang (2000) and Faheem et al.
(2003) suggested that the Prandtl slide failure is close
to the real basal failure mode of excavations. In the
paper, the Prandtls slide failure mode is adopted.
According to the distance between the hard stratum under the base of excavation and the excavation
base, the slide failure mode can be classified into two
typical modes as shown in Figure 3. Based on the
strength reduction finite element method, Goh (1990)
and Faheem et al. (2003) found that the shear strength
of clay-wall interface has slight influence on the base
stability. Here, we simply assume that the adhesion of
clay-wall interface is neglectable, which leads to the
simplicity in the upper bound analysis. The sliding surface consists of a 90 circular arc sandwiched between
two 45 isosceles wedges, an elastic wedge gjh and
plastic jik. The soil column efji acts as a surcharge.
According to the upper bound limit theorem, the
rate of external work should be equal to the rate of
internal energy dissipation in the system in a stability
problem. If a realistic, kinematically admissible sliding
mechanism is assumed, a reasonable collapse load can
be evaluated.
3.2
Calculation of basal stability
Assume the clay has a unit weight and a uniform

surcharge q is present adjacent to the excavation.
Case One: T Tc = B/ 2
From Figure 3(a), the rate of external work done by
(1) the weight of the soil columns efji and mnjg, (2) the
831
By equating dW to dE, the basal stability of excavation is
When the clay is homogeneous and the undrained

shear doesnt vary with the soil profile.
Figure 3. Velocity field based on assumed Prandtl sliding

mechanism.
Case Two: T < Tc

Similarly,
weight of soil in the two isosceles wedges jik and gjh,

and (3) the weight of soil in the radial zone jhk is
The total rate of internal energy dissipation from

(1) sliding along fi (2) sliding along ik and gh, and
(3) the radial shear combined with arc sliding in the
radial shear zone jhk is
where
When the clay is homogeneous and the undrained

shear doesnt vary with the soil profile.
PARAMETRIC STUDIES
The current analysis of basal stability can consider the

rotation of principal stress direction and anisotropic
832
Figure 5. Influence D/H on factor of safety.
Figure 4. Influence D/H on the factor of safety.
feature of clays. To study the effect of the strength

anisotropy, D/H and T /Tc to the basal stability of
deep excavation in clays, an infinitely long vertical deep excavation in soft clays is analyzed using
the aforementioned method. The unit weight of the
clay is = 18 kN/m3 , the undrained shear strength
is Suv (z) = 0.33v , the width of the excavation is
B = 15 m, the depth of the excavation is H = 12 m,
the penetration of the diaphragm is 6 m, 8 m, 10 m and
12 m, respectively.
4.1
Effects of strength anisotropy
Figure 4 shows the factor of safety of the basal stability of deep excavation varies linearly with the degree
of strength anisotropy of clay. In addition, the larger
degree of anisotropy, the smaller of safety factor. With
the same anisotropic ratio, increasing the penetration
of the diaphragm can not increase the basal safety
factor greatly.
4.2
FIELD CASE STUDIES
5.1 Boston case
Effects of D/H
Figure 5 indicates the safety factor of basal stability

increases with the increasing of D/H , which shows
the contribution of D/H to Fs . However, with the same
D/H , the anisotropic ratio is more influential than D/H
to the safety factor.
4.3
Figure 6. Influence T /Tc on the factor of safety.
Effects of T/Tc
Figure 6 indicates that presence of the bedrock close

to the base of the excavation increases Fs . This can be
explained that the size of the yielding zone is affected
since the displacement of the soil beneath and around
the excavation is restrained when the rigid stratum is
close to the base of excavation. However, with the same
T /Tc the factor of basal safety is decreased clearly with
the decreasing of anisotropic ratio.
Hashash and Whittle (1996) analyzed an deep excavation in Boston Blue Clays through finite element
method incorporating an advanced effective stress soil
model, MIT-E3 (Whittle and Kavvadas, 1994). The
study focus on an idealized (symmetric) plain strain
excavation geometry with half-width, B/2 = 20 m; the
depth of excavation is 10 m, 15 m, 22.5 m and 30 m
respectively; the penetration of the diaphragm is 2.5 m,
5 m, 17.5 m and 30 m respectively. The vertical effec
tive stress is expressed as v0
= 8.19z + 24.5, the
gravity of the clay is = 18.0 kPa. When the OCR
is 1.0, 2.0 and 4.0, the strength anisotropy ratio is 0.5,
0.48 and 0.43 respectively. In the current analysis the
author assumes the strength anisotropy ratio is about
k = 0.5 for BBC. According to the proposed method,
the factor of basal stability is 1.16 when the depth of
833
excavation is 10 m and penetration of diaphragm is 2.5

which quite agrees with the result (Hashash et al. 1996)
based on the finite element method incorporating with
the complicated MIT-E3 model.
5.2
Shanghai case
One failure case of excavation for a railway station

in Shanghai was reported by Jiang et al. (1997). The
width of the excavation is 23 m, the depth and the
length is 16.7 m and 600 m. The penetration of the
diaphragm is 13.3 m and 0.8 m thick. Soil parameters
concerned are K0 = 0.65, 1 k/ = 0.91, M = 0.818.
According to the formula for calculating the basal
stability recommended by the Shanghai Foundation
Design Code and the formula recommended by the
Shanghai Tunnel Engineering Design and Research
Institute, the safety factors are 2.4 and 1.4, respectively. However, the safety factor will reduce to 0.97
when the undrained shear strength is considered as
anisotropic and non-homogeneous.
CK0 UE triaxial tests. Comparisons between the result

of the current method and the numerical results presented by Hashash and Whittle (1996) for an excavation in normally consolidated Boston blue clay
demonstrate the ability of the proposed method to simulate the strength anisotropy and nonhomogeneity of
soils on the basal stability of deep excavations. Furthermore, this method also shows good accuracy in
back-analyzing the safety factor for two basal failure
cases in Shanghai and Taipei respectively. Conclusions
of the paper can be summarized as follows.
1 The necessities of considering the anisotropic feature of clays for the calculation of basal stability of deep excavations in soft clays have been
demonstrated.
2 Without taking into consideration of strength
anisotropy of clays, the safety factor against base
heave from the traditional methods is higher than
the real one.
3 The parameters used in the current method are
easily obtained by conventional lab tests (e.g. triaxial compression test and extension test or plain
compression and extension test). Additionally, the
method can consider the relationship between soil
strength and vertical effective consolidation stress.
5.3 Taipei case

Su et al. (1998) analyzed the base failure of an excavation in Taipei using an anisotropic strength formula
introduced by themselves. The dimension of this excavation is about 100 m long, 17.5 m to 25.8 m wide, and
13.45 m deep. The diaphragm wall retaining the excavation is 24.0 m deep and 0.7 m thick. Failure occurred
about two and a half hours after completion of the last
stage excavation and only two minutes were needed
before the entire internal bracing system collapsed.
The excavation site was located in a reclaimed land
along the Keelung River, which meanders through the
Taipei Basin. The top 8.7 m of the subsoil profile was
backfill and hydraulic fill materials. Underlying the
fills are a 2.0 m thick silty sand layer, a thick soft clay
layer ranging from Ground Level 10.7 to 44.7 m,
and a silty sand layer ranging from GL 44.7 m to
the bed rock located at GL 55.0 m. The distribu
tion of Suc and Sue with depth are Suc = 0.271vc
and

Sue = 0.189vc
, respectively, i.e. k = 0.7. The calculated safety factor is 1.10 which shows a possibility of
basal failure.
6
CONCLUSIONS
A method for analyzing the base stability of deep

excavations in anisotropic soft clay is presented in
this paper. It is derived from the upper bound theory
and strength reduction theory assuming the Prandtls
soil slide failure mode suitable for a critical basal
stability which can consider the anisotropy and nonlinear feature of clays. The undrained shear strength
and anisotropic ratio required for the method can
be determined from the conventional CK0 UC and
REFERENCES
Bishop, A.W. 1966. The strength of soils as engineering
materials. Geotechnique 16(2): 89130.
Bjerrum, L. & Eide, O. 1956. Stability of strutted excavations
in clay. Geotechnique 6: 3247.
Casarande, A. & Carillo, N. 1944. Shear failure of anisotropic
soil. J. of the Boston Society of Civil Engineers 31(4).
Chang, M.F. 2000. Basal stability analysis of braced cuts
in clay. Journal of Geotechnical and Geoenvironmental
Enigineering, ASCE, 126(3): 276279.
Chen, W.F. 1975. Limit analysis and soil plasticity. Elsevier
Scientific, Amsterdam.
Clough, G.W. & Hansen, L.A. 1981. Clay anisotropy and
braced wall behavior. J. Geotech. Eng. Div.,ASCE, 107(7):
893913.
Davis, E.H. & Christian, J.T. 1971. Bearing capacity of
anisotropic cohesive soil. J. Soil Mech. Found. Div.,
ASCE, 97(5): 753769.
Duncan, J.M. & Seed, B.H. 1966. Strength variation along
failure surfaces in clay. J. Soil Mech. Found. Div., ASCE,
92(9): 81104.
Faheem, H., Cai, F., Ugai, K. & Hagiwara, T. 2003. Twodimensional base stability of excavations in soft soils
using FEM. Comp. and Geotechnics 30(2): 141163.
Goh, A.T.C. 1990. Assessment of basal stability for braced
excavation systems using the finite element method.
Comp. and Geotechnics 10(4): 325338.
prediction for deep excavations in soft clay. Journal of
122(6): 474486.
834
Hu, Z.F., Zhou, J. & Yang, L.D. 2001. Study of subsoil stability under deep excavation. Chinese Journal of Civil
Engineering, 34(2): 8495.(in Chinese)
Jiang, H.W., Zhao, X.H. & Hang, B.L. 1997. Analysis of
heave-resistant stability for deep braced excavation in
soft clay under anisotropic condition. Chinese Journal of
Geotechnical Engineering, 19(1): 17.(in Chinese)
Ladd, C.C. 1973. Discussion. Main Session 4. In: Proc. 8th
ICSMFE, eds. Committee of 8th ICSMFE, Moscow, 4(2):
108115.
Lo, K.Y. 1965. Stability of slopes in anisotropic soils. Journal
of the soil mechanics and foundations division, ASCE,
91(SM): 85106.
Morgenstern, N.R. & Price, V.E. 1965. The analysis of the
stability of general slip surface. Geotechnique 15:7993.
Spencer, E. 1967. A method of analysis of the stability of embankments assuming parallel interslice forces.
Geotechnique 17:1126.
Su, S.F., Liao, H.J. & Lin, Y.H. 1998. Basal stability of deep
excavation in anisotropic soft clay. J. Geotechnical and
Geoenvironmental Eng., ASCE, 124(9): 809819.
Terzaghi, K. & Peck, R.B. 1948. Soil mechanics in engineering practice. Wiley, New York.
Ukritchon, B., Whittle, A.J. & Sloan, S.W. 2003. Undrained
stability of braced excavations in clay. J. Geotechnical and
Geoenvironmental Eng., ASCE, 129(8): 738755.
Wei, X. & Huang, M.S. 2006. Anisotropic bounding surface model for clays. Chinese Journal of Hydraulic
Engineering 37(7): 831837.(in Chinese)
Whittle, A.J., DeGroot, D.J., Ladd, C.C. & Seah, T.H. 1994.
Model prediction of the anisotropic behavior of Boston
Blue clay. Journal of Geotechnical Engineering, ASCE,
120(1): 199224.
Zou, G.D. 2004. Analysis of stability against upheaval of
deep excavation by an upper limit method. Rock and Soil
Mechanics, 25(12):18731878.(in Chinese)
835
Analytical two and three dimension models to assess stability and

deformation magnitude of underground excavations in soil
L. E. Sozio
Promon Engenharia, Sao Paulo, Brazil
ABSTRACT: Two and three dimensional elastic and plastic analytical solutions are presented, envisaging the
evaluation of stability and deformation of a tunnel or cavern. These solutions are based on thick wall cylinder
or thick wall sphere models, where inner and outer radius are defined according to tunnel or cavern geometry
and corresponding soil cover to ground surface. Radial body forces are introduced into the equations to emulate
gravity forces. Cohesive and frictional materials can be considered. The resulting model permits the calculation
of the plastic zone radius as a function of the support pressure acting at the tunnel or cavern inner surface and
obtain the critical support pressure. By adopting linear elastic behaviour at the elastic zone and no volume change
at the plastic zone it is possible to derive a complete convergence confinement curve. Radial displacement and
stress resulting from these models are compared with published case histories. Comparison is also made with a
known published elastic perfectly plastic plane strain model.
INTRODUCTION
support. The load on the support will depend on this

deformation.
A fundamental parameter in tunnel design is the allowable distance between tunnel face and the point where
support is effective. Support is generally effective upon
closure of the invert arch in NATM (New Austrian Tunnelling Method) tunnels (Fig. 1), or where segmental
lining is erected and grout is injected in non pressurized shield tunnels. In any case the tunnel Engineer
must anticipate if the opening will remain stable within
such length. If not, he should specify construction of
the invert arch closer to the face of the NATM tunnel,
or employ an earth pressure balanced type of shield. In
both cases he would be required to estimate the magnitude of soil deformation up to the activation of the
Figure 1. Tunnel length L, where support is not fully

effective.
SOME ASPECTS OF CURRENT DESIGN

PRACTICE
Notwithstanding the increasing availability of three

dimensional (3D) numerical computer program codes,
the predominant design practice of a tunnel lining is
still based on a two dimensional (2D) model. The relief
of ground stresses prior to the activation of the lining,
if accounted for, is generally simulated by a progressive softening of the soil within the excavated area.
The design engineer must rely either on his experience or on some published 3D analytical or numerical
analyses, which generally do not match the particular characteristics of his problem, to define how much
softening will be allowed prior to the lining activation.
If he allows little or no softening, the lining design
may be ultraconservative, particularly if the soil cover
above the tunnel is large. If he allows too much softening, the stresses on the lining may be underestimated,
and, if he defines a construction method allowing
for an excessive softening, his tunnel is in risk of a
collapse.
A comprehensive study into 3D effects at a tunnel
heading was performed by Negro (1988).
The analytical models presented herein are proposed to be used as a preliminary estimate of a tunnel
837
stress relief and corresponding soil displacements.

More sophisticated (and time consuming) analysis can
be performed by using 3D numerical models, at a
detailed design stage.
3
ELASTIC-PLASTIC ANALYTICAL
MODELS
These models are based on a thick wall sphere (3D

model) and on a thick wall cylinder (2D model) in
which part of the sphere or the cylinder is in elastic
state and the remainder in plastic state. The area of
the tunnel where support is considered either to be
not effective, or where a supporting pressure p can
be applied, is simulated as a sphere (Fig. 1) with an
inner radius Ri. The outer radius Ro (Fig. 2) simulates
the ground surface, where a surcharge pressure load
s can be considered. This concept has already been
used by Muhlhaus (1985) for a limit analysis where
the whole sphere is assumed to be in plastic state, but
no information is obtained on displacements prior to
collapse.
The analytical model was derived in a way that
either a 3D or a 2D analysis can be performed. The
3D analysis is the objective of this concept model.
Although less interesting, the 2D model can be used
to roughly evaluate average lining load vs. soil displacement, after lining installation, bearing in mind
that the axi-symmetry model condition does not allow
for bending of the lining.
In the plastic zone of the sphere or cylinder the stress
state is found by combining the differential equation
of equilibrium (in polar coordinates) and the Mohr
Coulomb criterion.
In the elastic zone of the sphere or cylinder the stress
state is found by combining the differential equation
of equilibrium, the stress strain relationships and the
compatibility strain displacement relationships. Refer
to Figure 3 for notation and equations.
Radial body forces are included in the equilibrium equations of both the plastic and elastic zones
to emulate gravity forces.
The boundary between the elastic and plastic zones

(Rp in Figure 2) is determined by equating the elastic
stresses to the Mohr Coulomb criterion. Note that for
unstable configurations no solution is found for that
equation. On the other extreme, if Rp Ri, then the
whole sphere or cylinder is in elastic state.
The radial displacement at the elastic-plastic boundary is obtained from the elastic equation (9) of Figure 3,
which is derived from the differential equation of equilibrium, the stress strain relationships and the compatibility strain displacement relationships, applied
to the elastic zone, and considering the previously
determined radial stress at elastic-plastic boundary.
For displacements in the plastic zone it is neccessary
to consider the plastic deformation behaviour. No volume change was adopted for this analysis, as imposed
on equation (10). This equation should be modified if
dilating or contracting behaviour is to be modelled.
If Ro , equation (9) will reproduce the classical elastic expression for radial displacement due to a
cylindrical or spherical opening in an infinite elastic
medium.
Frictionless materials (undrained analysis) have a
particular set of equations to define plastic radius,
since for friction angle equal to zero, parameter M = 0
(equation 2 of Fig. 3) make the general equations indeterminate. A check has been made by adopting a very
small value of M , for instance M = 1e5 , and the general and undrained analyses give practically the same
results.
The equations of Figure 3 can be inserted into a
programmable pocket calculator (such as the HP32s,
which contains an algorithm that solves equation (7a)
or (7b) = equation (8)) to enable rapid assessment of
a given case.
4
LIMITATIONS
These models were developed to give a preliminary

insight into stability and deformation behaviour of
underground excavations, with a particular interest
in tunnel headings. The main limitations that should
always be considered are listed below:
4.1 Radial simmetry
Figure 2. Thick Wall Cylinder (2D) or Sphere (3D).
The models imply in radial simmetry, thus any shape

of tunnel or sphere, and ground surface will have to
be approximated to a cylinder or sphere. One type
of approximation required when considering a tunnel heading is the change of shape of the unsupported
length of tunnel into a sphere, as shown in Figure 1.
Some geometrical situations may arise for which a
great distortion occurs, and one of particular interest is
a long heading of a shallow tunnel. If such a condition
is to be analysed, is advisable to complement the evaluation by conservatively checking the results of a 2D
838
Figure 3. 2D and 3D elastic-plastic model equations.
839
analysis, for an infinite cylinder of inner radius equal

to the long heading radius.
4.2
Constitutive model
Soil is assumed to be homogeneous, isotropic and linear elastic with Mohr Coulomb plastic criterion and
no volume change. Soils stress strain behaviour is in
general not linear, and this would affect mainly the
displacements estimate.
4.3
In situ stresses
Radial simmetry implies in radial body forces, other

than vertical gravity. The effect of a horizontal to vertical in situ stress ratio different than one is not possible
to be considered in these models.
Also, the effect of a flow net resulting from ground
water table above the tunnel axis cannot be accounted
for, except for the simplified assumption of a descending vertical flow net, with hydraulic gradient equal to
one, thus implying in a saturated (in lieu of submerged)
unit weight of the soil. In general a flow net towards
the tunnel heading represents risk of a hydraulic piping
failure, and should be avoided by vertical or horizontal
drains, compressed air, or a slurry type pressure.
4.4
For the 3D analysis the inner radius was defined

to equate the unsupported length tunnel volume, additionaly considering the tunnel face as a half sphere of
diameter equal to the tunnel diameter,
Consolidation
Settlements due to consolidation are not considered

in this model, analyses are either fully undrained or
drained.
4.5
Figure 4. Example of convergence confinement curve.
Local collapse
Local Collapse as defined by Davis et ali (1980), is

not included in these models.
5 APPLICATIONS AND COMPARISONS
WITH OBSERVED DATA
Figure 4 shows an example application of the 2D and
3D models for a tunnel, considering a given length L
between tunnel face and effective support activation.
The settlement of the tunnel roof on activation of the
lining is estimated as the displacement of the sphere
inner surface, which is the intersection of the 3D model
convergence curve with the abcissa (Fig. 4). At this
point the lining is activated.
The final pressure on the lining and crown settlement are estimated from the intersection of the tunnel
lining stress displacement characteristic curve with the
2D model convergence curve.
An analysis can be made on the sensitivity of the
unsupported length; the crown settlement increases
with increase of this length up to the point where the
whole sphere is in plastic state, resulting in collapse
conditions.
where D is the tunnel Diameter.

Eight tunnel case histories were analysed in terms of
the models presented herein, see Table 1 for references
and Table 2 for parameters and results.
The tunnel geometry and soil strength and deformability parameters were inferred from the respective
case history papers. Drained parameters were used for
sandy soil cases and fully undrained parameters were
used for clayey soil cases. In the absence of better
information, the undrained deformability moduli of
clays were taken as 600 times their undrained shear
strength. The measured settlement figures were taken
on activation of the lining, although this was not significantly different from the final settlements on all
cases. For the Washington D.C. shield case it appears
that soil movements were partially restrained to move
freely at the rear of the shield steel hood. For this case
the unsupported length in the analysis was adopted
slightly smaller than the actual distance from face to
the segmental lining installation.
The stresses on the lining were measured in the
Edmonton cases (case 7 and case 8, see Table 1) and
their authors published the idealized soil convergence
confinement curve plus the carachteristic curves of
two types of linings, concrete segments and steel ribs
with wood lagging. The 2D model was applied to this
case and the resulting curve is shown in Figure 5. The
convergence confinement curve obtained from the 3D
model to simulate the unsupported length of the tunnel
is also shown, for the concrete lining condition.
840
A reasonable approximation to the measured loads

and / or settlements for this and the other cases can be
observed.
A comparison was also made with the model presented by Mair & Taylor (1993), for the unloading of
plane strain cylindrical cavity, for linear elastic, perfectly plastic undrained soil behaviour. Curves from
both models and for a specific set of parameters are
shown in the graph of Figure 6. It was found that
the conformity between these models is affected by
the soil cover to tunnel radius ratio. In general the ratio
between radial displacements of each model increases
as the tunnel becomes shallower. This could be due to
the fact that the Mair & Taylor model is applicable to
an infinite medium condition.
6
The elastic plastic 3D analytical model presented

herein is meant to be used as an auxiliary tool to enable
assessment of tunnel heading stability conditions and
deformation magnitude prior to installation of support.
CONCLUSIONS
Despite the significant evolution on numerical computer program codes, two dimensional tunnel analyses
are still predominant in the current design practice.
Consequently the 3D stress state in the zone of the tunnel heading is often either ignored or unsatisfactorily
simulated.
Table 1.
Figure 5. Convergence confinement curves, Edmonton

cases.
Case histories and references.
Case
Reference
1
2
3
4
5
6
7
8
Frankfurt, NATM, Baulos 25 Cording (1976)

Frankfurt, Shield, Fahrgasse Cording (1976)
Washington DC, F2A Line 1 Cording (1976)
Heathrow Cargo Cording (1976)
Green Park Jubilee Line Attewell (1974)
Sao Paulo E-W Line Sozio (1978)
Edmonton, concrete Eisenstein (1979)
Edmonton, steel ribs Eisenstein (1979)
Figure 6. Comparison between Mair & Taylor and 2D

model.
Table 2. Application of 3D model vs. observed data.

Case
Diameter m
Lm
Ri m
Ro m
kN/m3
c kPa
E MPa
Rp m
w mm
calculated
w mm
observed
6.5
3.0
3.44
14.6
19.0
55
0
33
0.5
6.68
42
6.5
8.0
4.32
12.4
18.5
15
30
35
0.25
9.07
69
5.5
5.0
3.38
20.1
20.0
15
35
80
0.25
5.20
23
10.9
5.0
5.77
13.3
19.0
100
0
60
0.50
6.11
11
4.1
4.0
2.59
29.3
19.0
150
0
90
0.50
4.33
20
6.1
7.0
3.98
20.0
19.0
25
33
60
0.25
5.20
22
2.6
6.0
2.03
26.9
20.0
110
0
66
0.50
4.52
38
2.6
11.0
2.44
26.9
20.0
110
0
66
0.50
5.33
43
45
70
21
14
17
25
33
38
841
The 2D model permits an estimate on the stress level

to be expected on the support.
Comparisons with some published case histories
and with a known elastic perfect plastic analytical model are presented, no significant discrepancies
being observed.
A series of limitations have to be considered when
employing these models, as outlined above (Heading
4). Therefore these models are not to be used as a
sole design tool, but as complementary elements to
the accepted tunnel design procedures and standards.
REFERENCES
Attewell, P. & Farmer, I. 1974. Ground deformations resulting from shield tunnelling in London clay. Canadian
Geotechnical Journal 11.
Cording, E. & Hansmire, W. 1976. Displacements around
tunnels in soil. U.S. Department of Transportation Report
TST-76T-22.
Davis, E., Gunn, M., Mair, R. & Seneviratne, H. 1980. The

stability of shallow tunnels and underground openings in
cohesive material. Geotechnique, 30.
Eisenstein, A., El-Nahas, F. & Thomson, S. 1979. Pressure displacement relations in two systems of tunnel
linings. 6th Soil Mechanics and Foundation Engineering
Conference, Lima.
Mair, R.J. & Taylor, R.N. 1993. Prediction of clay behaviour
around tunnels using plasticity solutions. Predictive Soil
Mechanics, Proc. Wroth Memorial Symp., Oxford, U.K.
Muhlhaus, H.B. 1985. Lower bound solutions for circular
tunnels in two and three dimensions. Rock Mechanics &
Rock Engineering, 18.
Negro, A. 1988. Design of shallow tunnels in soft ground.
PhD. Thesis, University of Alberta, Canada.
Sozio, L.E. 1978. Settlements in a Sao Paulo shield tunnel.
Tunnels & Tunnelling, vol. 10 n 7.
842
Dynamic response of saturated silty clay around a tunnel under subway

vibration loading in Shanghai
Y.Q. Tang & Z.D. Cui
X. Zhang
ABSTRACT: This paper studies the saturated soft clay around the tunnel between Jingansi Station and Jiangsu
Station of Shanghai subway Line No.2. The continuous dynamic monitoring is conducted by means of embedded
earth pressure piezometers and pore piezometers around the tunnel at different locations and different depths.
The response frequency and stress amplitude of the saturated soft clay are studied with the distance from the
tunnel due to the subway vibration loading. A formula is proposed for the attenuation of the dynamic response
of the soil. The distance of influence and the amplitude of the dynamic response are calculated and the influence
on the surrounding buildings under the subway vibration load is predicted, which offers valuable references for
the design, the construction and the safe operation of the subway.
INTRODUCTION
The subway is indispensable as a safe, comfortable and

high speed transport vehicle, in modern cities, but in
the course of its running, the vibration problem should
not be ignored. Some models (Alabi, 1992; Lipen and
Chigarev, 1998; Sheng, et al., 1999; Jones, et al., 2000)
were used to analyze the ground vibration under the
train loading on the railway. Pan (1995), Hirokazu
(2001) and Xie (2002) studied the dynamic response
of the railway system, but they focused only on the
response of the upper structure of the foundation and
did not study the dynamic response of the soil. Up to
now, only a few researchers have studied the dynamic
response of the saturated silty clay around the subway tunnel. Yet, the long-term vibration loading of the
subway has resulted in large deformation of the saturated silty clay (Chen, et al., 2002; Wang, et al., 2003).
According to monitored data, a large deformation of
the axis of the subway tunnel and the ground settlement
had occurred in some sections of the tunnel of Shanghai subway Line No.1 (Lin, et al., 2000) and affected
the normal operation of the subway. The settlement and
deformation originally began with the change of pore
water pressure and soil stress (Tang, et al., 2003; Tang,
et al., 2005). So it is important to study the dynamic
response of the soil around the subway tunnel for the

safe operation of the subway.
As the soil around the subway tunnel in Shanghai
is mainly the saturated silty clay, this paper focuses on
it under the subway vibration loading. The response
frequency of the saturated soft clay and the law of the
amplitude of stress response changing with the depth
and the distance away from the tunnel are all studied
under the subway vibration loading. The result can
offer a valuable reference to the design, construction
and the safe operation of the subway.
2
DYNAMIC MONITORING
In order to study the influence caused by the subway

vibration loading on the saturated soft clay around the
tunnel, field test and monitoring are conducted in this
research. The site is selected between Jingansi Station
and Jiangsu Road Station. The dynamic monitoring
system is adopted for field monitoring and its sampling frequency can reach 200 Hz and its precision is
0.1 kPa. It can fully reflect the soil response around
the tunnel due to the subway vibration loading. The
dynamic monitoring system consists of a resistance
sensor, a dynamic strain amplifier, a data selector
843
Strain gauge
electric signal
resistance
sensor
strain
amplifier
1.30m
2.70m
1
2
electric signal
4
digital signal
data
selector
computer
Figure 1. Scheme of dynamic monitoring system.
BH3
BH4
1800
Subway
People Square
2000
~
~
20.0m
~
~
~
~
~ ~
~ ~
~ ~
~ ~
~ ~ ~
~ ~ ~ ~
8.5m
~
1.8m
~
R3m
11.5m
~ ~
Tunnel
13.5m
~
~
~ ~ ~
~
~ ~ ~
~
20.0m
Figure 3. Soil profile and embedded instruments.

Note: layer No.1 is mixed soils; layer No.2 is brown yellow
silty clay; layer No.3 is gray muddy silty clay; layer No.4 is
gray mucky clay; layer No.5 r is gray silty clay.
Zhong Shan Park

Monitoring
Room
~ ~
~ ~ ~
~ ~ ~ ~
~
7.80m
~ ~ ~
~ ~ ~ ~
~
~ ~ ~ ~
~ ~
~ ~ ~
~ ~ ~ ~ ~
~ ~ ~
~
~ ~ ~
~
~ ~ ~ ~ 18.0m
~
silty clay of layer No.4 and the earth pressure piezometers and pore piezometers are located at the depths of
8.5 m, 11.5 m and 13.5 m, respectively, in layer No.4,
to monitor the response characteristic of the vibration
for the subway running.
BH5
BH2
3000
3
BH1
15000
15000
Figure 2. Layout of boreholes (unit: mm).
and a computer. The system can record all the sampling data collected by the computer in real time. The
dynamic monitoring system is shown in Figure 1.
Figure 2 shows the layout of boreholes at the site.
In the plane, there are five boreholes, each 110 mm
in diameter, parallel and vertical to the subway tunnel, respectively. The distance between the site and
Jingansi Station is 210 m. Boreholes BH3, BH4 and
BH5 are parallel to the tunnel axis only 1.8 m away
from the outside of the segment of the subway tunnel
and the distance between them is 15.0 m. Boreholes
BH1, BH2, and BH3 are vertical to the tunnel axis. In
order to study the attenuation of the effect on the soil
around the tunnel with the increasing distance under
the subway vibration loading, the distance between
boreholes BH1 and BH2 is 3 m and that between boreholes BH2 and BH3 is 2 m, so that there is a step-up
course.
Figure 3 shows the distribution of strata and instruments. In the section, the subway tunnel lies in gray
RESPONSE FREQUENCY OF SOIL
The train of Shanghai subway has six carriages and

its whole length is 139.46 m. Its normal running speed
is 60 km/h and the break at each station is 3040 s.
The monitoring site is near the subway station. The
train will decrease its speed when running into the station. According to statistics, the speed is generally at
3040 km/h when the train passes through the monitoring site, and the time is 1216 s in general. The
interval of the subway train crossing the monitoring
site is unequal, generally 36 min and 34 min during
rush hours.
There are two groups of wheels in every carriage
of the subway train, and the subway vibration loading
is transferred to the soil outside the tunnel segments
through the system of wheels-segments. When the
subway train comes near the monitoring site, the vibration loading is produced and spread in waves, and
the soil can be induced to respond. In Figure 4, the
horizontal axis is time and the vertical axis is stress
response, and the wave can be seen clearly when the
subway train runs across the monitoring site. Owing
to the different loading of each carriage (the number of passengers being always different), the response
amplitude is also different. The response frequency of
soil is also different because of the different interval of
the groups of wheels. There are two types of response
844
Figure 4. Wave of the soil response.
High frequency
Low frequency
Figure 6. Attenuation of horizontal stress vertical to tunnel

axis.
Frequency (Hz)
2.5
taken into consideration, the corresponding measure

should be adopted to avoid the equality of monitoring frequencies and natural frequency in the design of
subway tunnel. This can result in resonance, causing
different kinds of calamity, such as the instability of
the subway tunnel, the fissure of the tunnel segments,
and so on.
2
1.5
1
0.5
0
No.1
No.2
No.3
No.4
No.5
Hole number
DYNAMIC RESPONSE OF SOIL
4.1 Attenuation of dynamic response of soil vertical

to the axis of the subway tunnel
Figure 5. The response frequency of soil for the five

boreholes.
frequency: the high frequency fh and low frequency

fl . The high frequency is produced mainly by the rear
wheels of the first carriage and the front wheels of the
second carriage. The distance between them is short,
so the response time of soil is also short. When the
two groups of wheels run across the monitoring site,
the high frequency is produced. The low frequency is
produced mainly by the front and rear wheels of the
same carriage and the distance is longer, so it is lower.
The vibration response of soil can be seen clearly when
every group of wheels runs across the monitoring site,
as shown in the figure.
By continuous dynamic field monitoring, plenty of
data are obtained. All the waves are analyzed and compared and the typical response wave of the saturated
soft clay is obtained. The response frequency of soil
(silty clay of layer No.4) of every borehole is obtained
by collecting, analyzing and arranging its data. The
statistic values are 2.42.6 Hz and 0.40.6 Hz for
the high and low frequency, respectively, as shown in
Figure 5.
When these two kinds of response frequency and
the natural frequency of soft clay in Shanghai area are
The test and research on the attenuation of vibration

of the train on the ground with the increasing distance from origin of vibration were mainly focused
on the transverse direction. From Chens test (Chen,
et al., 1998), the vibration almost attenuated to zero
at 25 m away from the rail of the railway, and had no
effect beyond 25 m. The same idea was proved by other
tests, and only the distance of attenuation was different for different vibration loadings and conditions. The
subway tunnel is embedded in the semi-infinite soil
body, and the value of attenuation is larger than that
of the ground. Boreholes BH1, BH2 and BH3 are horizontally ardistanced at different locations vertical to
the tunnel axis and are 6.8 m, 3.8 m and 1.8 m away
from the edge of the subway tunnel, respectively. In
order to find the attenuation law more clearly and try
to avoid the interference by other factors, the soil at
13.5 m depth is chosen for study. The monitored data
are averaged and the result is given in Figure 6.
According to the statistic result of large amounts
of monitored data, the relationship between the soil
response and the distance vertical to the subway tunnel
axis is given in Formula (1):
845
where p (kPa) is the value of dynamic response

of soil when the subway is running; K0 (kPa) is the
value of dynamic response of soil at the edge of tunnel; K1 (kPa/m) is the first order coefficient of the
attenuation of dynamical response with the distance;
K2 (kPa/m) is the second order coefficient of the attenuation of dynamical response with the distance; x is the
distance away from the edge of the subway tunnel.
According to field monitored data by fitting Formula (1), the coefficients are obtained: K0 = 1.3526,
K1 = 0.0321 and K2 = 0.0077.
Substitution of coefficients into Formula (1) yields:
Formula (2) is giving the dynamic response attenuation law of horizontal soil vertical to the tunnel
axis with distance in this field monitoring, where the
distance of x is 011.33 m.
Using Formula (2), the distance of influence and
the amplitude of dynamic response can be calculated.
The influence on the surrounding buildings under the
subway vibration loading can be predicted. It can offer
a theoretical reference to the design and construction
of the building near the subway tunnel.
It can be seen from Figure 6 that the attenuation of
the subway vibrating loading takes on a certain rule
with the increasing distance. The influence distance is
11.33 m obtained from Formula (2). The result is quite
different from that of the dynamic railway loading on
the ground. In addition to the factor of soil, the main
reason is that the shear modulus and the damping of
soil increase with the increasing depth. This makes the
dynamic loading attenuate rapidly to vanish. Along the
subway, the rails are joined with tunnel segments and
the rigidity is much larger than that of the soil vertical
to the subway.
4.2
Dynamic response law of soil at different

depths
Sensors are embedded at the top, the middle and at the

bottom of the tunnel. The early static monitoring data
indicate that the soil pressure approximately takes on
the linear increase with the increasing depth. In order
to find the relationship between the stress amplitude
and depth, the statistic data at the rush hour are selected
for analysis, as shown in Figure 7.
It can be seen from Figure 7 that under the subway
vibration loading, the maximum changing amplitudes
of the stress response of soil are 0.23 kPa at 8.5 m,
0.70 kPa at 11.5 m and 1.15 kPa at 13.5 m. The change
of stress amplitude is approximately linear with depth.
Moreover, the stress amplitude at the rush hour in the
morning is larger than that at the rush hour in the
evening, which is also larger than that at noon. This
indicates that the number of passengers is the largest
Figure 7. Amplitude of soil pressure response with depth.
in the morning. This should be treated as a reference to

design and calculate the worst case in the construction.
CONCLUSIONS
1 The attenuation of the subway vibration loading vertical to the axis of subway tunnel takes on a certain
rule with the increasing distance. According to the
statistic results of large amounts of monitored data,
the formula for the attenuation of dynamic response
of the horizontal soil vertical to the axis of the subway tunnel is obtained with the increasing distance.
Using this formula, the influence distance and the
amplitude of dynamic response can be calculated.
The influence on the surrounding building under
the subway vibration loading can be predicted. It
offers a theoretical reference to the design and
construction of buildings near the subway tunnel.
2 The change of stress amplitude of soil is approximately linear with depth when subway trains are
running across. Moreover, the stress amplitude is
the largest during the rush hour in the morning. It
indicates that the flow of passengers is the largest
during the rush hour in the morning when the soil is
in the worst condition. This value should be adopted
as reference for the design and construction of the
subway tunnel.
3 According to the two kinds of response frequency
by the field monitoring of the soil around the tunnel, which are produced by the subway vibration
loading and matched with the natural frequency of
soft clay in Shanghai area, the corresponding measurement should be adopted in the design of the
subway tunnel in order to avoid the response frequency equal to the natural frequency. It could result
in resonance causing kinds of calamity, such as the
instability of the subway tunnel, the fissure of the
tunnel segments, etc.
846
ACKNOWLEDGEMENTS
This work are supported by the research grant
(40372124) from National Natural Science Foundation of China, Shanghai Key Subject (Geotechnical Engineering) Foundation and Shanghai Leading
Academic Discipline Project(Project Number: B308).
REFERENCES
Alabi, B. 1992. A parametric study on some aspects of
ground-borne vibrations due to rail traffic. Journal of
Sound and Vibration 153(1): 7787.
Chen,Y.M., Chen, R.P. & Lu, S. 2002. Several soil mechanics
problems in metro construction and operation on soft clay
foundation. In:Proceedings of Seminar on Metro Construction and Environmenta1 Geotechnical Engineering.
Hangzhou: (in Chinese).
Hirokazu, T., Shuhei, S. & Xie, W.P. 2001. Train track-ground
dynamics due to high speed moving source and ground
vibration transmission. Journal of Structure Mechanics
and Earthquake Engineering 682(7): 299309.
Jones, C.J., Sheng, X. & Petyt, M. 2000. Simulations of
ground vibration from a moving harmonic load on a
railway track. Journal of Sound and Vibration 231(3):
739751.
Ling, Y.G., Liao, S.M. & Liu, G.B. Discussion of Influencing factors on axial deformation of subway tunnel.
Underground Space 20(4): 264267(in Chinese).
Lipen, A.B. & Chigarev, A.V. 1998. The displacements in an

elastic half-space when a load moves along a beam lying
on its surface. Journal of Applied Maths Mechanics 62(5):
791796.
Pan, S.C., Li, D.W. & Xie, Z.G. 1995. The discussion of
the influencing on environment by subway train vibration.
Journal ofVibration and Shock 14(4): 2934(in Chinese).
Sheng, X., Jones, C. & Peryt, M. 1999. Ground vibration
generated by a harmonic load acting on a railway track.
Journal of Sound and Vibration 225(1): 328.
Tang,Y.Q., Hang,Y. &Ye, W.M. 2003. Critical dynamic stress
ratio and dynamic strain analysis of soil around the tunnel
under subway traffic loading. Journal of Rock Mechanics
and Engineering 22(9): 15661570(in Chinese).
Tang, Y.Q., Zhang, X., Zhou, N.Q. & Huang, Y. 2005. Microscopic study of saturated soft clays behavior under cyclic
loading. Journal of Tongji University 33(5): 626630(in
Chinese).
Wang, C.J., Ji, M.X. & Chen, Y.M. 2003. Subsequent settlement of saturated soft clay ground induced by train. In:
Proceedings of the 9th Conference on Soil Mechanics and
Geotechnical Engineering. Beijing: Tsinghua University
Press (in Chinese).
Xie, W.P., Hu, J.W. & Xu, J. 2002. Dynamic response of trackground systems under high speed moving load. Journal
of Rock Mechanics and Engineering 21(7):10751078(in
Chinese).
847
Lateral responses of piles due to excavation-induced soil movements

C.R. Zhang, M.S. Huang & F.Y. Liang
ABSTRACT: Lateral soil movements induced by excavation of a deep foundation pit may adversely affect
nearby pile foundations. In this paper, a simple analysis method is proposed for computing lateral responses
of passive pile groups subject to excavation induced lateral soil movement. Based on a two-stage method, the
Winkler model is adopted for simulating the pile-soil interaction, combined with finite difference method in the
case of multi-layered soils. A specified free-field soil movement profile is used as input. Then, the governing
equation for a pile group is obtained considering the shielding effect in pile groups by the simplified Mindlins
equation. Comparisons are made between the observed behavior of centrifuge model tests and those computed
by the proposed method. The present method can in general give a satisfactory prediction of the lateral response
of passive pile groups. However, the major limitation is the assumption of linear elastic soil springs, which
provides only an upper bound estimate.
INTRODUCTION
Deep excavations for basements and other underground facilities are unavoidable in big cities. The
lateral soil movements resulting from excavations will
impose additional bending moments and deflections
on nearby piles, which may lead to structural distress or
failure (Pan et al. 2002, Goh et al. 2003). Developing
reliable and simple methods to estimate the behavior of piles next to excavations is urgent in practical
engineering.
Available methods of analysis can be broadly classified into complete three-dimensional method and
two-stage method. The former, which is carried out
by a finite element analysis, can consider complex
pile-soil interaction and the whole construction process (Goh et al. 2002, Miao et al. 2005). However, it is
computationally expensive and the accuracy depends
on the accuracy of the constitutive soil models, which
are currently under a stage of calibration with results
from physical tests (Juirnarongrit & Ashford 2006).
Furthermore, it is more suited to obtain a benchmark
solution or to obtain solutions of detailed analysis for
final design, rather than as a preliminary routine design
tool (Kitiyodom & Matsumoto 2002).
Comparatively, the simplified two-stage method
appears to be a more attractive choice. The known freefield soil movement is a prerequisite (Goh et al. 1997).
As a wealth of experience accumulations on estimation of the soil deformation resulting from excavation
have been obtained by engineers, the two-stage method
can give a satisfactory result to guide construction and

design. In the works of Poulos (Poulos & Chen 1997,
Chen & Poulos 1997, 1999), a combined finite element method and boundary element method was used
to analyse piles adjacent to an excavation. However,
most of their works are focused on single passive piles
and the surrounding soil is modelled as an elastic continuum. Issues such as group effect due to pile-soil-pile
interaction and the effect of non-homogeneous soils
are not well-understood. Further research is clearly
required.
This paper describes a simplified two-stage numerical procedure for analyzing response of piles in
group subjected to excavation-induced lateral soil
movements. A numerical model, considering of nonhomogeneous soils, is achieved using the finite difference method and the concept of shielding effects
is introduced to analyse the pile-soil-pile interaction.
The assessment is performed by comparing the results
from the presented method with those from centrifuge
model tests and the predicted results of centrifuge tests
by Leung (Leung et al. 2000, Leung et al. 2003).
2 ANALYSIS METHOD
According to the two-stage method, the analysis can
be decomposed into two components (Poulos & Chen
1997). First, the free-field soil movement (without the
presence of piles in the substratum) is obtained by
849
measurement or calculation. Second, the acquired soil

movement is imposed on a nearby pile to calculate its
response. The flexural bending of the pile is modelled
by an elastic beam while the complex phenomenon
of pile-soil interaction is modeled by linear elastic soil
springs based on the Winkler model. The lateral deflection equation for a single pile is formulated. Then,
considering the restriction of soil movement due to
pile-soil-pile interaction, the shielding effect between
two piles is imposed using simplified Mindlins elastic
solution for a lateral point load in an elastic half-space.
At last the response of group piles is obtained with
superposition theorem.
2.1
of subgrade reaction kz to the elastic parameter Es and

s of the soil mass, as follow in:
where Ep Ip is pile rigidity.

Then, the governing differential equation of single
pile is given by:
in which Ut (z) is the lateral deflection of pile caused

by excavation; hx (z) is the free-field soil movements
due to excavation; and is written as
Free-field soil movement induced by the

construction of deep foundation pit
Analysis methods for estimating free-field soil movement include an empirical method, a finite element
method and an analytical method. As a means to verify
the validity of the two-stage method, supposed freefield soil movement or that from in-situ measurement
can be adopted. Bigot et al. (1982) clearly showed that
the displacement-based method of analysis provides
very good predictions of bending moment profiles
and pile deflections if the measured free-field soil displacement is used as input, or if an accurate prediction
of soil movement can be made.
2.2 Analysis of a single pile

As an approximate method, the nonlinear springs
which represent the actual pile-soil interaction and tension cracks developed around piles are not taken into
account. In other words, the following hypotheses are
adopted:
1 the pile is represented by an elastic beam based on
a Winkler subgrade reaction model;
2 the complex phenomenon of pile-soil interaction
is modeled using linear elastic soil springs and no
crack appears between pile and surrounding soils;
3 the effect of axial load on the pile is ignored.
The equation (3) can be solved either by finite difference method or by analytical method. The analytical
method has been undertaken by the authors before, and
can only be applied for homogeneous soils. In order to
consider the influence of non-homogeneous soils, the
numerical finite difference method is adopted in this
paper.
The total pile length L is divided into n cells and the
length of each cell is , with a node in each cell, which
is 0,1,. . . n 1, n. In addition to n real cells, there are
four additional imaginary cells (two at each end of the
pile and additional nodes of 2, 1, n + 1, n + 2) to
implement boundary conditions.
The basic differential equation (3) can be written in
the finite difference form for any real cell i, as
The different boundary conditions at the top and tip

of the pile provide four different additional equations
(Poulos & Davis 1990).
With free-head pile exerted load H and bending
moment M, equations are as follow:
The linear elastic soil spring is represented through

a modulus of subgrade reaction, which is defined as
where kz has the units of force/length2 , p = soil reaction per unit-length of pile, in unit of force/length,
y = the relative displacement between pile and surrounding soils. Vesic (1961) analyzed an infinite
horizontal beam in an elastic foundation and compared the results with those obtained by the use of
subgrade-reaction theory, which related the modulus
With fixed pile head, the equations are
When deflection and rotation of pile head is possible,

the equation is changed corresponding to the relevant
boundary condition.
850
kzi Us1,i from Winkler subgrade model in the case of

active piles. Based on Mindlins equation, the corresponding soil displacement in arbitrary node j at the
position of pile 2(without the presence of pile 2) due
to load at node i is
Figure 1. Computation model for two passive piles.
Similarly, the equations for cell n 1, n can also be

acquired, based on their boundary conditions.
Substituting the equations of boundary conditions
into equation for nodes 0, 1, n 1 and n, the n + 1
unknowns, which represent the deflection of pile
in n + 1 nodes, can be solved by the whole system of n + 1 simultaneous equations. The rotation,
shear force and bending moment of the pile can
be obtained from deflection based on the theory of
material mechanics.
2.3
where R21 = s2 + (zj zi )2 , R22 = s2 + (zj + zi )2 , zi and

zj is the depth of cell i, j, x is the distance between
two piles in the direction of lateral soil deflection. The
coefficient ij can be therefore written as
With the superposition theorem, the lateral soil shielding displacement in any node j at the position of
pile 2 is
Interaction between two piles
In general, piles do not follow exactly the free-field

soil movement at the pile location and the soil profile is also altered by pile. The hindered free-field soil
movement affects nearby piles, the amount of which
depending on the relative stiffness between pile and
surrounding soils. This is called shielding effect.
The problem of interaction between two piles is
depicted in Figure 1, where two piles with pile spacing
s are represented. By discretizing each pile into n cells
with a node in each cell, the free-field soil movement
in arbitrary node i at the position of pile 1 is hs1,i and the
deflection of pile 1 caused is Up1,i . The corresponding
lateral soil deflection is Us1,i , where Up1,i is identical
to Us1,i based on compatibility of lateral displacement
between pile and soil. The lateral soil displacement in
node i due to shielding effect of pile 1 is then expressed
as
The lateral equilibrium equation of pile 2 due to

shielding effect of pile 1 is written as
where Ut21 (z) is the corresponding lateral deflection

of pile 2. The equation can also be solved by finite
difference method in the same way.
2.4
Pile group analysis
For the usual case of a group of n piles, the total

response of a single pile in the group(i.e. the total displacement and the total internal force) is written as the
sum of two components:
1 displacement due to excavation-induced free-field
soil movement.
2 shielding displacement resulting from pile-to-pile
interaction (shielding effect) which decreases the
response of single pile caused by excavation.
The corresponding lateral soil shielding displacement in any node j at the position of pile 2 due to soil
displacement in node i at the position of pile 1, is
where ij is the attenuation function of the lateral

shielding movement based on the simplified Mindlins
equation as follows (Poulos & Davis 1990). Suppose
the load in arbitrary node i at the position of pile 1 is
Considering the arbitrary pile i in the group, the

total displacement of the pile head could be obtained
by superposition
851
where Utii represents the response of pile i due

to excavation-induced free-field soil movement, and
Utij (i = j) represents the shielding displacement of pile
i due to the existence of pile j.
3 VERIFICATION BY COMPARISON WITH

CENTRIFUGE MODEL TESTS AND FEM
ANAYSIS
3.1
Figure 2. Lateral soil movement induced by excavation.
Experiments
Only very limited field data are available so far,

especially in the case of pile groups. Leung et al.
(2000, 2003) have published a series of centrifuge
model tests on unstrutted deep excavation in dense
sand and its influence on an adjacent single pile
and pile group foundation behind the retaining wall.
Figure 3. Comparison of free-head pile response for pile located 1 m behind wall.
Figure 4. Comparison of free-head pile response for pile located 3 m behind wall.
852
Due to approximate assumption about linear elastic

soil springs in the present simplified method, cases
in which soil experienced failure are not taken into
account. The predicted results by Leung in the case of
single pile with finite element method are also shown
herein.
3.2 Analysis of single piles
Two tests for single free-head pile located at 1 and 3 m
behind the retaining wall and two tests for restraint
head (fixed deflection-free rotation head and fixed
deflection-fixed rotation head) at 3 m behind the
retaining wall are presented here. The prototype square
pile has a width of 0.63 m and a length of 12.5 m with
Ep Ip = 220 103 kNm. The free field soil displacement profiles with depth at the location of the pile
are instrumented and shown in Figure 2. Es = 6z (in

MPa, z is the depth below ground surface, in m) is
applied for the analysis as suggested by Leung et al.
(2000). Figures 34 show comparisons between the
measured and predicted bending moments and deflection profiles along a free-head pile located at 1, 3 m
from the wall. Figures 56 give the results for the case
of head restraint. FE analysis by Leung et al. (2000)
gives a relatively better estimation than the present
method, especially at the position of the maximum
bending moment. The reason maybe the assumption
of linear elastic soil springs along the pile. The soil
around the long pile up to a depth less than 4 times the
pile diameter may have reached limited soil pressures,
even experiencing very small displacement (Pan et al.
2002). In Leungs works, the ultimate soil pressures
acting on pile were introduced to consider this effect,
Figure 5. Comparison of fixed (deflection)-free(rotation) head pile response.
Figure 6. Comparison of fixed (deflection)-fixed(rotation) head pile response.
853
while they are not taken into account in the presented method. However, the simplified method still
gives reasonable estimation with little computational
effort and it can be used with some confidence in the
preliminary design.
3.3 Analysis of pile group
17 group-pile tests with different number of piles and
different configures are shown in Leung et al. (2003).
Due to the limited length of this paper, only 2 typical
pile groups are described here. The prototype square
pile has a width of 0.48 m and a length of 12.5 m with
Ep Ip = 240 103 KNm2 . The pile cap with a thickness of 0.55 m is placed above the ground, which can
be treated as rigid cap. Pile groups with four piles
in free-head and capped-head are described here, the
configuration of which are shown in Figure 7. The
predicted results by the present simplified method are
compared with the data from centrifuge tests in Figures 810. The calculated results provide reasonable
approximation to the centrifuge tests data for freehead and capped-head pile groups. The discrepancy
between the predicted and measured bending moment
profiles is seem to be small along the upper portion
of the pile, while it is relatively large along the lower
portion of piles. However, deflections of capped-head
piles show the tendency that the front pile is dragged
back by the rear file through connection of pile cap.
Figure 7. Configuration of pile group.
CONCLUSIONS
This paper presents a numerical analysis with finite

difference method for studying the behavior of piles
subjected to excavation-induced lateral soil movement
in non-homogeneous soils. Response of a single pile
is determined by imposing the known free-field soil
movement profile to the passive pile. The Mindlins
Figure 8. Comparison of free head pile group.
Figure 9. Comparison of front pile in capped head pile group.
854
Figure 10. Comparison of rear pile in capped head pile group.
equation is employed to calculate the shielding effect

of passive pile groups due to pile-soil-pile interaction.
Comparisons with centrifuge model tests confirm that
the method provides reliable estimates in simple way.
However, the major limitation of the method is the
assumption of linear elastic soil springs, which provides only an upper bound estimate of the maximum
bending moments and pile deflections. Hence, the
analysis considering the nonlinear effect is still needed
in order to decisively assess the responses of piles due
to excavation-induced lateral soil movements.
ACKNOWLEDGMENTS
This work was supported by Shanghai Leading Academic Discipline Project, Project Number: B308.
REFERENCES
Bigot, G., Bourges, F. & Frank, R. 1982. Etude Experimental
Dun Pieu Soumis Aux Poussees Laterales Du Sol. Revue
Francaise de Geotechnique 18: 2947.
Chen, L.T. & Poulos, H.G. 1997. Piles subjected to lateral
soil movements. Journal of Geotechnical and Geoenvironmental Engineering, ASCE, 123(9): 802811.
Goh, A.T., Teh, C.I. & Wong, K S. 1997. Analysis of
piles subjected to embankment induced lateral soil movements. Journal of Geotechnical and Geoenvironmental
Engineering, ASCE, 123(9): 792801.
Goh, A.T.C., Wong, K.S., Teh, C.I. & Wen, D. 2003.
Pile response adjacent to braced excavation. Journal of
129(4):
Juirnarongrit, T. & Ashford, S.A. 2006. Soil-pile response
to blast-induced lateral spreading. II:analysis and assessment of the py method. Journal of Geotechnical and
Geoenvironmental Engineering 132(2):163172.
Kitiyodom, P. & Matsumoto, T. 2003. A simplified analysis

method for piled raft foundations in non-homogeneous
soils. Int. J. Numer. Anal. Meth. Geomech. 27:
85109.
Leung, C.F., Chow, Y.K. & Shen, R.F. 2000. Behavior of
pile subjected to excavation-induced soil movement. Journal of Geotechnical and Geoenvironmental Engineering
126(11): 947954.
Leung, C.F., Lim, J.K., Shen, R.F. & Chow,Y.K. 2003. Behavior of Pile Groups Subject to Excavation-Induced Soil
Movement. Journal of Geotechnical and Geoenvironmental Engineering 129(1):5865.
Miao, L.F., Goh, A.T.C., Wong, K.S. & Teh, C.I. 2006.
Three-dimensional finite element analyses of passive
pile behavior. Int. J. Numer. Anal. Meth. Geomech. 30:
599613.
Pan, J.L., Goh, A.T.C., Wong, K.S. & Teh, C.I. 2002. Ultimate
soil pressure for piles subjected to lateral soil movements. Journal of Geotechnical and Geoenvironmental
Pan, J.L., Goh,A.T.C., Wong, K.S. & Selby,A.R. 2002. Threedimensional analysis of single pile response to lateral soil
movements. Int. J. Numer. Anal. Meth. Geomech. 26:747
758.
Poulos, H.G. & Chen, L.T. 1997. Pile response due to
excavation-induced lateral soil movement. Journal of
Geotechnical and geoenvironmental engineering 123(2):
9499.
Poulos, H.G. & Davis, E.H. 1990. Pile foundation analysis and design. Florida: Pobert E. Krieger Publishing
Company.
Stewart, D.P. 1999. Analysis of piles subjected to embankment induced lateral soil movements. Journal of Geotechnical and Geoenvironmental Engineering, ASCE, May:
425426.
Vesic, A.S. 1961. Bending of beam resting on isotropic elastic solid. Journal of. Engineering Mechanics Division,
ASCE, 87(2): 3553.
855
Elastic-plastic analysis for surrounding rock of pressure tunnel with lining

based on material nonlinear softening
L.M. Zhang & Z.Q. Wang
College of Science, Qingdao Technological University, Qingdao, P.R. China
ABSTRACT: The elastic-nonlinear softening-residual plastic surrounding rock mode1 is analyzed. According
to the total strain theory, the relationship between equivalent stress and equivalent strain is deduced from uniaxial
compression of practical rock. The relational expressions are related with triaxial stress ( , z , and r ) and
triaxial strain ( , z , and r ) of surrounding rock. The mechanism of load bearing and acting relation between
surrounding rock and support are studied. Stress distribution and displacement of broken zone, plastic zone and
elastic zone of tunnel are presented out. The ultimate bearing capacity of surrounding rock is given. The critical
pressure leading to yield firstly of surounding rock that caused by inner pressure is introduced. The obtained
results are more closely to actual values of surrounding rock. It is pointed out that there are obvious limitations
in Kastener formula, which is based on ideal elastic-plastic model. Analysis shows that the ideal plastic model
and the brittle model are special cases of the proposed solution.
INTRODUCTION
Kastners solution is often used in elastic-plastic analysis for surrounding rock of circular tunnel. There
are obvious limitations in Kastners formula, which
is based on ideal elastic-plastic constitutive model.
This leads to the Kastners solution is far away to
coresponding actual values in surounding rock of tunnel. Following along the path of pioneered by Kastner,
researchers such as Ma (1995, 1996), Jiang and Zheng
(1996, 1997), Ma (1998, 1999),Yu (2002), Fan (2004),
Ren and Zhang (2001) and Pan and Wang (2004)
published different solutions for surrounding rocks of
circular tunnel. However, these solutions are resticted
to very simple material models, such as simple linear
relationship between stress-strain. They are of limited practical value. This study successfully gets the
stress distribution laws of surrounding rock plastic
and broken zone of tunnel according to the total strain
theory. The relationship between equivalent stress and
equivalent strain is deduced from practical rock. The
relational expressions are related with triaxial stress
( , z , and r ) and triaxial strain ( , z , and r ) of
surrounding rock.
2
ELASTIC-PLASTIC ANALYSIS FOR

SURROUNDING ROCK
between its internal and external pressures. Where a

and b are respectively the inner and outer radius of
tunnel, P0 and Pa are pressures acting on the inner and
outer surfaces of tunnel, and R is the radius of the interface of elastic and plastic zones. Surrounding rock may
be generally divided into broken, plastic and elastic
zones on the basis of their states, as shown in Fig. 1.The
surrounding rock within the elastic zone is in an elastic state, within the plastic zone, in a strain softening
state, and within the broken zone, in a residual-strength
state. So the surrounding rock within the broken zone
is the direct object of tunnel support.
2.1
Constitutive model of plastic zone
The tunnel may be simplified as an axisymmetrical

plane strain problem. Substituting r = r3 , z = 0 and
i = c into the interface of elastic and plastic zones,
the equivalent strain is defined by (Zheng, 1988)
If the volumetric strain of softening zone equals to

zero, we can obtain
Fig. 1 shows the geometric model condition of a tunnel

in a plane strain state subjected to a pressure difference
857
of surrounding rock keep constant propotion, ie

z : : r = 0:1: (1). So it may be simplified as simple loading condition. According to the total strain
theory (Zheng, 1988), the relationship between equivalent stress and equivalent strain can be deduced
from Eq. (4)
The relational expressions are related with triaxial

stress ( , z , and r ) and triaxial strain ( , z , and r )
of surrounding rock.
2.2 Stresses in the plastic zone
Its corresponding mechanical equilibrium equation is
Figure 1. Model of tunnel.
Substituting (r )r = r2 = p2 into Eq. (3) and Eq. (5)

and using the Eq. (6), the stresses in the plastic zone
can be obtained as follows
Figure 2. Equivalent stress-equivalent strain.
The equivalent stress is given by
2.3 Stresses in the elastic zone

The stresses in the elastic zone may be expressed by
The constitutive model of uniaxial compression is

calculated from (Guo, 2004)
where c is a constant, i refers to the stress corresponded to peak strength, i refers to the strain
corresponded to peak strength. The constant c may
be determined by the results of a set of uniaxial tests.
The softening section on equivalent i i curve can
be plotted as Fig. 2 shown. We find that the assumed
rock mode1 agrees well with practical rocks.
The ultimate bearing capacity of surrounding rock
in complex stress state is analyzed in the following parts. We consider that the strain component
Substituting the bounding condition on the interface

of elastic and plastic zones (e )r = r3 + (re )r = r3 =
p
p
2p0 = ( )r = r3 + (r )r = r3 into Eq. (7) and Eq. (8),
we get
858
The plastic zone has the support effect to surrounding rock. Eq. (9) shows that for a given p0 , surrounding
rock can be balanced by itself through adjusting the
plastic zone. So it is also named the equilibrium equation of surrounding rock. Surrounding rock without
support has ultimate bearing capacity. If r3 ,
we get the ultimate bearing capacity of surrounding
rock. In practical, the surrounding rock is collaped as
r3 . However, it gives us the theoretical result.
The ultimate bearing capacity of surrounding rock in
practical can not be larger than the theoretical result
of surrounding rock.
Substituting Eq. (7) into Eq. (13), we have
2.4 Deformation in the plastic zone
According to the elastic-plastic theory, the toatal strain
of plastic zone can be calculated by the following
formula (Zheng, 1988)
Combining Eq. (14), we can solve Eq. (15) and get the
following formula
The elastic strain of plastic zone is defined by
The plastic strain of plastic zone is defined by
where
The total strain of plastic zone can be expressed by
where Ec refers to elastic modulus of surrounding rock,

Gc refers to shear modulus, c refers to Poissions
radio, refers to plastic function. The plastic functio is zero in elastic deformation. Using Eq. (13),
the deformation on the interface of elastic and plastic
zones can be calculated from
859
Substituting r = r2 into Eq. (16), we have
2.5
Combining Eq. (22) and Eq. (20), we get
Stresses and deformation in the broken zone
The broken zone cannot bear the tangential stress,

so the tangential stress is zero. Its corresponding
mechanical equilibrium equation is
3
SUBMISSION OF MATERIAL TO THE

EDITOR
The lining can be considered as thick-wall cylinder in inner pressure pa and outer pressure p1 . The
deformation of lining may be expressed by
The deformation in the broken zone may be expressed

by
Combining Eq. (23) and Eq. (24), we get
Combining Eq. (17), we can solve Eq. (21) and get the
deformation formula
Combining Eq. (25), Eq. (20) and Eq. (9), we get
where f =
(1 20 )r1
.
E0
860
the characteristic of the strength dropping after softening of surrounding rocks plasticity. It is suitable
for analyzing the stress of plastic zone of wall rock
openings.
(2) If the rock does not enter plastic state, but has
broken zone, substituting r3 = r2 into Eq. (25), we find
If the rock does not enter broken state, but has

plastic zone, substituting r2 = r1 into Eq. (25), we find
Figure 3. Ideal plastic model of rock.
If the rock does not have plastic zone and broken

zone, substituting r3 = r2 , r2 = r1 into Eq. (25), we find
Figure 4. Brittle model of rock.
Using Eq. (26), we may get r3 . Substituting r3 into

Eq. (25), we find the stress formula of lining as follows
(3) Substituting r3 = r2 into Eq. (26), we get the formula of critical pressure leading to yield firstly for
surounding rock that caused by inner pressure
DISCUSSION OF RESULTS
(1) According to Eq. (5), we know

If c = 0, we get = c . The curve of strength after
yield changes into horizon line, as shown in Fig. 3.
Which means the equation stands for ideal plastic
material. This suggests that the ideal plastic model,
such as Fenners solution and Kastners solution is
special cases of the proposed solution.
If c , we get = 0. The curve of strength after
yield changes into zero, as shown in Fig. 4. Which
means the equation stands for brittle material. This
shows that the brittle model is special case of the
proposed solution.
If 0 < c < , the equation stands for the behavior
of post-failure strength. Different integration constant
c means different behavior of strength after yield,
as shown in Fig. 2. The equation betterly describes
CALCULATION EXAMPLE AND

EXTENSION OF ITS APPLICATION
Typical cross-section of a pressure tunnel is shown in

Fig. 1. The design length of tunnel is 150 m with inner
diameter of a = 3.0 m, outer diameter of r1 = 3.5 m.
The mechanics parameters of rock and lining can
be gotten by test, E0 = Ec = 30 MPa, Ed = 60 MPa,
c = u0 = ud = 0.25, d = 50 MPa, r2 = 3.5 m, p0 =
1.3 MPa, pa = 2 MPa.
According to Eq. (31), we get the critical pressure
pcr
a = 1.67 MPa. As the inner pressure pa is greater
than the critical pressure pcr
a , plastic zone occurs. Then
we get r3 = 5.01 m from Eq. (26). The loosen range
of surrounding rock is obtained by Ultrasonic tests
to be 5.15 m. It is very close to the theory result to
be 5.29 m from Eq. (26). Table 1 shows stresses of
different positon of lining from Eq. (29).
861
Table 1.
REFERENCES
Stresses of different position of lining.
r/m
3.00
3.25
3.50
rc (kPa)

(kPa)
c
1.96
1.92
1.74
1.62
1.47
1.37
CONCLUSIONS
Here we may draw the following conclusions.

1 The elastic-nonlinear softening-residual plastic surrounding rock model is analyzed. According to the
total strain theory, the relationship between equivalent stress and equivalent strain is deduced from
uniaxial compression of practical rock, which is
related with triaxial stress ( , z , and r ) and triaxial strain ( , z , and r ) of surrounding rock.
Stress distribution laws of different position of
surrounding rock, the mechanism of load bearing
and acting relation between surounding rock and
support are studied. Analysis shows that the ideal
plastic model and the brittle model are special cases
of the proposed solution.
2 Different radial stresses of the interface under different conditions, such as elstic-plastic condition,
elastic-broken conditions are obtained. The ultimate bearing capacity of surrounding rock is given.
The critical pressure leading to yield firstly for
surounding rock caused by inner pressure is also
obtained.
Fan, H. & Yu, M.H. 2004. An analytic solution of elastoplastic pressure tunnel considering material softening and
dilatancy. Engineering Mechanics 21(5):1624
Guo, Z.H. 2004. The strength and constitutive model of
conctrete. China Architecture and building Press, Beijing
Jiang, M.J. & Sheng, Z.J. 1996. On expansion of cylindrical
cavity with linear softening and shear dilatation behaviour.
Journal of Rock Mechanics and Engineering 16(6):
550557
Ma, N.J. 1996. A new analysis on ground pressures around
openings. Journal of Rock Mechanics and Engineering
15(1): 8489
Ma, N.J. 1999. Multilinear strength attenuation model of rock
body and plastic area of wall rock of openings. Journal of
Metal Mine 9: 1012
Ma, G.W., Iwasaki, S. & Miyamoto, Y. 1998. Plastic limit
analysis of circular plates with respect to unified yield
criterion. Int. J. Solids & Structure 43: 1137115
Ma, G.W., Iwasaki, S. & Miyamoto,Y. 1999. Dynamic plastic
behavior of circular plats using the unified yield criterion.
Int. J. Solids & Structure 36: 32573275
Pan, Y. & Wang, Z.Q. 2004. Research on relationshp of
load-displacement for cavern surrounding rock with strain
nonlinear softeningng. Journal of Rock Mechanics and
Engineering 25(10):15151521
Ren, Q.W. & Zhang, H.C. 2001. A modification of fenner
formula. Journal of Hohal University 29(6): 109111
Yu, M.H. 2002. Advances in strength theory of materials
under complex stress state in the 20th century. Applied
Mechanics Reviews 55(3): 169218
Zheng, Y.T. 1998. Fundamentals of elastic-plastic-sticky
theory of rockmechanics. Coal industry Press, Beijing
862
Modification of key parameters of longitudinal equivalent model for

shield tunnel
W. Zhu & X.Q. Kou
Geotechnical Research Institute, Hohai University, Nanjing, Jiangsu, P.R. China
X.C. Zhong
College of Civil Engineering, Hohai University, Nanjing, Jiangsu, P.R. China
Z.G. Huang
Suzhou Traffic Design and Research Institute, Suzhou, Jiangsu, P.R. China
ABSTRACT: The shield tunnel actually is a slender construction which has many weak spots in the longitudinal
direction. Because of the character, the longitudinal non-uniform deformation of the tunnel would happen easily.
With increasing of shield tunnel in China, the problem of the longitudinal non-uniform deformation has become
attracted. The equivalent continuous model which has clear concept and could be calculated simply has been used
widely in projects to analyze the longitudinal performance of the tunnel. The shell-spring model which bases on
the three-dimensional FEM was introduced to analyze and modify the error of the equivalent continuous model
which was due to the increasing of the width of linings segment. The suggested range of the effective ratio of
the longitudinal rigidity in common situations of the shield tunnel was induced.
INTRODUCTION
With the development of municipal works in china,

the shield tunnels which are mainly used in soft ground
have been in a widespread availability. The layer which
has been traveled by the shield tunnel has become
more and more sophisticated. So the problem that
longitudinal non-homogeneous deformation occurs on
the tunnel lining because of its traveling sophisticated
layers has become attracted.
Now there are mainly two kinds of exiting longitudinal calculation models which are equivalent continuous model and longitudinal beam-spring model to
analysis the longitudinal mechanical property of the
shield tunnel linings. The equivalent continuous model
simplifies the tunnel a continuous beam which is on
the elastic foundation in the longitudinal direction and
recognizes that the cross section of the tunnel lining is
uniform. At the same time, it considers that the deformation of the tunnel in the longitudinal direction is
continuous at the longitudinal joints. The influence of
the longitudinal joints of rings is simplified by the
way of equivalent rigidity. The equivalent continuous
model has the clear conception and it is easy to be
calculated, so the model has been used widely in the
actual projects.
With the advancement in the design and construction

experience, there is a trend to increase the width of the
segmental linings. For example, in Japan, the width
of the lining segment was 0.8 0.9 m in past, it was
developed to 1.0 m in 1975, and the width of the tunnel linings used in Gulf of Tokyo was indeed up to
1.5 m; in China, the width was developed from 1.0 m
for shanghai subway to 1.5 m for Guangzhou No.2
subway, and the width of 1.2 m has been commonly
adopted in Guangzhou subway No.1 line, Nanjing subway, Shenzhen subway and Beijing subway No.5 line.
The segment ring width of the tunnel which traverses
the Yangzi River in Nanjing was 2.0 m. The number
of joints can be reduced by 20% by increasing the
width of the lining segment. The production cost can
be reduced, and the waterproof quality of the tunnel
can be improved because of reduction of the number
of joints.
The equivalent continuous model discussed above
assumes that the joints of ring have the influence on
mechanical performance of rings segment along the
whole width. This model considers that the delivery
of the deformation and internal force of linings along
the longitudinal direction is continuous and homogeneous. Actually with the increasing of the width of the
segment ring of linings, the shear force produced by
863
the bolts of the joints leads to the increases in bending

stress which are concentrated on the edge of the ring.
The delivery of the deformation and internal force of
linings along the longitudinal direction is inhomogeneous especially at the joints of rings because of the
concentration of the stress. The equivalent continuous
model which is the 2D model isnt able to reflect the
complexity of the deformation of the linings increasingly while the width of segment linings increasing.
Thus it is a need to modify the equivalent continuous
model and make it suitable to the development of the
shield tunnel.
Huang & Zang (2003) considered the influence of
the prestress of the longitudinal bolts on the longitudinal rigidity of linings, and modified the equivalent
continuous model.
Liao (2002) improved the equivalent continuous
model, considered that the influence area of joints of
rings is finite and not the whole ring length and suggested that the effective ratio of longitudinal bending
rigidity of the Shanghai subway was in the range of
1/71/5.
Huang & Xu (2005) suggested the values of the
effective ratio of longitudinal and transversal rigidity of shield linings in Shanghai area though indoor
modeling tests.
In this study, the 3D numerical simulation called
the shell-spring model which could reflect the actually mechanical performance of the segmental linings
properly was introduced to modify the equivalent continuous model. In this model, the segments of the
shield tunnel lining are simulated by the association
of shell elements, the joints of the segments are simulated by the spring-elements, and effects of the soil
support are realized by the soil-spring elements.
Figure 1. Typical section of tunnel lining.
Figure 2. Local coordinates and global coordinates.
2.2 Segmental joint element

2 THREE-DIMENSIONAL MODELLING OF
SHIELD TUNNEL LINING
Figure 1 shows a typical section of shield tunnel lining. In the shield tunnel construction, the segmental
linings of each ring are jointed together by the bolts
in transverse direction of tunnel. Similarly, two consecutive rings are connected together by the bolts in
the longitudinal direction of tunnel. In the numerical
modeling, two coordinate systems were used: global
coordinates (x , y , z ) as shown in Figure 1 and local
coordinates (x, y, z) which were used to describe the
stresses and deformations for a single element.
2.1
Shell element
In this study each segmental lining was modeled by a

mesh of three-dimensional shell elements. Each shell
element has four nodes. Each node has six degrees of
freedom, which are axial and angular displacements
in x, y and z direction.
Each ring of tunnel lining is composed of several

segmental linings, which are connected together by
the bolts in the transverse direction of tunnel. The segmental joint element was used to simulate the joints
between the segmental linings. Zhu and Tao (1998)
have put forward the beam-joint discontinuous element to simulate the mechanical behavior of linings
and joints in both transverse and longitudinal direction
of tunnel. They adopted the non-linear numerical analysis through the concept of Goodman element (1968).
In their model the joint element exhibits tensile resistance to simulate the bolt connection. Their numerical
result shows that the beam-joint discontinuous model
was suitable for simulating the non-linearity of joint
and discontinuous deformation of linings.
In this study, the concept of Goodman element was
used together with the three-dimensional joint element. The axial, shear and rotation of the segmental
joint element in the local coordinates system are
described by axial stiffness (Ksx ), shear stiffness in
864
radial and tangential shear in the linings, the tension

and compression along the longitudinal direction of
tunnel were also considered.
The axial, shear in radial and tangential direction of
the longitudinal joint element in the local coordinates
system (x, y and z) are described by axial stiffness
(Krz ), shear stiffness in radial direction (Kry ), shear
stiffness in tangential direction (Krx ), respectively.
Each longitudinal joint element has two coincidence
nodes. The discontinuity of joint between two rings is
denoted by the relative displacement of two nodes in
the local coordinates systems. Each node of the longitudinal joint element has three degrees of freedom
which are axial displacement in x y and z direction.
The radial and tangential shear, and axial force in the
longitudinal direction of tunnel are considered in each
node. Compressive stiffness is assumed to be infinite.
Tensile stiffness is the same as the stiffness of bolt,
k = EA/L, where E is Youngs modulus of bolt, A is
cross-section area, L is length.
Figure 3. Segmental joint element.
2.4
Soil-spring element
The interaction between the segmental tunnel linings

and the surrounding ground was modeled by the soil
spring in this study. In this paper, the radial and tangential spring was used to simulate the radial and
tangential force acting on the segmental linings from
the ground. The springs were arranged over the whole
circular section of the tunnel lining.
3
Figure 4. Longitudinal joint element.
radial direction (Ksy ), shear stiffness in tangential

direction (Ksz ) and rotational stiffness (K ), respectively. Each segmental joint element has two coincidence nodes. The discontinuity of the joint between
two segmental linings is denoted by the relative displacement of two nodes in the local coordinates system
x, y and z direction. Each node of the segmental joint
element has four element has four degrees of freedom,
which are axial displacement in x, y, z direction and
angular displacement in z direction.
2.3 Longitudinal joint element
The shear model was used to simulate the reinforcing
effects in the longitudinal direction of tunnel induced
by the staggered arrangement of linings in the beamjoint discontinuous model. The shear model include
radial and tangential shear of the linings. In this study,
the beam-joint discontinuous element was extended to
the three-dimensional joint element. In addition to the
NUMERICAL SIMULATION AND

ANALYSIS
The numerical simulation was used to calculate the

longitudinal rigidity of the shield tunnel linings. In
this study, the effect of the soil didnt need to be considered. So the soil-spring elements in the simulation
system were neglected. The simulated model is shown
in figure 5. There are four rings in the model and
boundary condition is selected to be one side fixed
and one side free. The outside and inside diameter
is 6.2 m and 5.5 m. the linings were constructed by
C50 concrete. The youngs modulus of concrete is
3.45 104 N/mm2 (see Table 1).
The arrangement of segmental linings is stagger
joint. Figure 5 shows that there are six segmental linings in each section. Segment A B and K have a central
angle of 67.5 , 68.0 and 21.5 , respectively. The linings are arranged in sequence and in stagger in the
longitudinal direction of tunnel, as shown in Figure 5.
There are 18 bolts which connected two consecutive
rings of linings in the longitudinal direction of tunnel.
For analyzing the longitudinal banding rigidity of
the linings, the model of numerical simulation (shown
in Figure 6) was adopted. In this study, the linings
is one end fix and one end free for loading. The
865
Figure 7. Equivalent continuous model for shield lining.
The mainly model parameters in numerical simulation are shown below.

Outside diameter of the tunnel: D = 6.2 m; inside
diameter: d = 5.5 m;
The thickness of the lining: t = 350 mm;
The diameter of the bolts: d = 30 mm;
The length of the bolts: l = 400 mm;
The number of the bolts: n = 17;
Youngs modulus of the bolts: Es = 2.1
105 N/mm2 ;
Poissons ratio of the bolts: = 0.3;
Yield stress of the bolts: [y ] = 6.4 105 N/mm2 ;
Breaking stress of the bolts: [f ] = 6.4
105 N/mm2
The grade of the concrete: C50
Youngs modulus: Ec = 3.45 104 N/mm2 ;
Poissons ratio: = 0.3.
Figure 5. Arrangement of segmental linings in first and

second sections.
Table 1. Influence of the lining segment width to the
longitudinal equivalent bending rigidity.
Width of lining segment/m
1.4
1.8
Analytical solution/GPa
Numerical simulation/GPa
72.43
100.8
100.64
159.7
128.48
205.9
3.2 The influence of the lining width to longitudinal

rigidity of tunnel lining
The equivalent continuous model simplifies the tunnel
a continuous beam which is on the elastic foundation in
the longitudinal direction and recognizes that the cross
section of the tunnel lining is uniform. At the same
time, it considers that the deformation of the tunnel in the longitudinal direction is continuous at the
longitudinal joints. The influence of the longitudinal
joints is simplified by the way of equivalent rigidity
(Figure 7).
The equivalent banding rigidity of the tunnel:
Figure 6. Simulation model.
displacement of the free end of lining would be

observed.
3.1
The effective rate of equivalent banding rigidity:
Model parameters
The model parameters used in the segmental and longitudinal joint element were chosen based on the national
and international shield tunneling experience and are
summarized. For the stiffness of segmental joint, the
values are the total stiffness of one ring. In the numerical simulation, each ring is further divided into four
portions in the longitudinal direction of tunnel as
shown in Figure 4.
could be solved with this formulation:
866
Where Kj the elastic stiffness rigidity of all longitudinal bolts, Kj = nkj

kj the elastic stiffness rigidity of one longitudinal
bolt
Ic inertia moment of the section
Ic =
[D4 (D t)4 ]
64
Ec Youngs modulus of the linings
Ac cross-section area of the linings
the position angle of the neutral axis.
Figure 8 shows the relationship between the average vertical displacements of the observed points at
the free end which calculated by the simulation model
based on shell-spring model and the loads when the
width of linings is 1 m, 1.4 m and 1.8 m. It could be
concluded that the displacements which is calculated
by the simulation model are generally lower than the
analysis solution which is calculated by the equivalent
continuous model.
3.3 Analysis of simulation results
As the Table 1 shown, the equivalent rigidity of
the equivalent continuous model is lower than the
numerical values. Actually, the influence area of the
longitudinal joints to the whole linings deformation is finite. And the simplified assumption of the
model which considers the influence is at the whole
ring of lining is not correct. The equivalent banding
rigidity of the equivalent continuance model should
be lower than the actual values. Besides that, the
influence of the arrangement of segmental linings
to the longitudinal performance of the tunnel isnt
considered by the equivalent continuous model and
the influence of that should not be ignored when
the internal force delivers along the longitudinal linings. The local deformation at the joints of rings also
could influence the delivery of the endogen force.
This phenomenon is just so-called local arrangement
effect.
With the lining width developed, the difference of
longitudinal equivalent rigidity between the simulation
model and the equivalent continuous model became
bigger (Figure 8 & Table 2).
It was also approved that with the lining segment
width increasing the influence of the joints of rings
to the whole rings performance became reduced. And
this change above can not be reflected by the equivalent
continuance model which generally adds the influence
of the joints of rings to the whole rings performance
as the width of lining developed. And, the error of the
equivalent continuous model would increase (Table 2)
while the width of linings increases.
As the Table 2 shown, the effective ratio of the longitudinal rigidity has been increased when the width of
Figure 8. Relationship between the load and the average

vertical displacement of the observed points at the free end.
Table 2. Influence of the lining segment width to the

effective ratio of the longitudinal rigidity.
Width of lining segment/m
1.4
1.8
Equivalent continuous
model
Numerical simulation
Difference between the
above two models (%)
1/12.14
1/8.74
1/6.85
1/8.73
39.1
1/5.51
58.6
1/4.27
60.4
lining increased. Actually, with the number of joints

of rings and its influence decreased the longitudinal performance of the segmental lining tunnel has
become more similar to the homogeneous tunnel.
867
CONCLUSIONS
1 The equivalent bending rigidity which is used in

the analytical solution of the equivalent continuous
model is lower than the actual values. Because of
the assumption of the model, the calculated method
of the equivalent bending rigidity in the equivalent
continuous model couldnt reflect the influence of
the segmental arrangement effect which generally
causes the delivering discontinuity of the internal
force of the linings at the joints of rings.
2 With the width of lining increasing, the error of
the equivalent bending rigidity in the classical
equivalent continuous model would increase.
3 When the width of linings is the general range of
1.0 m 1.8 m used in the shield tunnel engineering
in China, the effective ratio of the longitudinal bending rigidity of the common situations of the subway
could be in the range of 1/8.73 1/4.27.
Hashimoto, T., Zhu, H. H., & Nagaya, J. 1994. A New Model

for Simulating the Behavior of Segments in Shield Tunnel.
Proceeding of the 49th Annual Conference of the JSCE,
12421243.
Huang, H.W., & Xu, L. 2006. Study on Transverse Effective Rigidity Ratio of Shield Tunnels. Chinese Journal of
Geotechnical Engineering, 28(1):1118. [in Chinese]
Huang, H.W., & Zang, X.L. 2002. Research and Analysis
on Longitudinal Deformation Characteristics of Shield
Tunnel. Underground Space, 22(3):244250. [in Chinese]
ITA Working Group 2, 2000, International tunneling association, guidelines for the design of shield tunnel lining.
Tunneling and Underground Space Technology, 15(3),
pp.303331.
Lee, K.M., Hou, X.Y., Ge, X.W., & Tang Y. 2001. An
analytical solution for a jointed shield-driven tunnel lining. International Journal for Numerical and Analytical
Methods in Geomechanics, 25(4):365390.
Zhu, H.H., Yang, L.D., & Chen, Q.J. 1996. Two design
models for segment joint of lining system in shield
tunnel. Engineering Mechanics, Supplement, 395399.
[in Chinese]
REFERENCES
Goodman, R.E., Taylor, R.L., & Brekke, T.L. 1968. A model
for the mechanics of jointed rock. Journal of Soil Mechanics and Foundation Division, ASCE, 94: 11211129.
868
Author Index
Ahmadi-adli, M. 295, 301

Ai, X.Q. 499
Alhieb, M. 649
Amorosi, A. 615
Antiga, A. 365
Arslan, U. 459
Assis, A.P. 519
Augarde, C.E. 785
Avdarsolkyzy, S. 751
Dijkstra, J. 621
Ding, G.Y. 545
Ding, W.Q. 525
Ding, Y.H. 455
Du, J.H. 643
Baimakhan, A.R. 751

Baimakhan, R.B. 751
Bakker, K.J. 243, 249, 255
Bao, H.L. 757
Bao, W. 817
Bao, W.Y. 215
Bao, X.H. 507
Batali, L. 187
Bezuijen, A. 3, 243, 249, 255,
261, 281, 349, 357
Boldini, D. 615
Bolton, M.D. 15
Broere, W. 621
Burghignoli, A. 627
Farias, M.M. 519

Fillibeck, J. 275
Finno, R.J. 87
Cai, Y.M. 465

Caporaletti, P. 627
Chan, Y. 471
Chang, G.M. 133
Chao, H.C. 67
Cheah, C.K. 447
Chen, C. 513
Chen, D.C. 579
Chen, J. 761
Cheng, L.C. 799
Chi, Y. 491
Chin, C.T. 67
Chiorboli, M. 365
Chissolucombe, I. 519
Cho, G.C. 45, 717
Choi, J.H. 669
Chuay, H.Y. 393
Chung, S.S. 567
Cong, L. 235
Cui, Z.D. 769, 843
Danaev, N.T. 751
Dashdorj, S. 751
Date, K. 635
Deng, A.Z.G. 269
Dias, D. 373
Eclaircy-Caudron, S. 373
Emeriault, F. 77, 207, 689
Erridaoui, A. 385
Gafar, K. 281
Gao, W.J. 607
Geraldni, Y.D.S. 655
Gong, Q.M. 381
Guatteri, G. 287
Guiloux, A. 385
Guo, B.H. 531, 799
Guo, W. 817
Gutierrez, M. 537
Hajialilue-Bonab, M. 295, 301
Han, C. 513
Han, X. 331, 545, 725
Hashimoto, A. 173, 405
Hashimoto, T. 99, 173, 307
He, C. 313
Hoefsloot, F.J.M. 357, 775
Hong, C. 781
Hou, J. 551
Hrustinec, L. 325
Hsi, J. 141
Hsiung, B.C.B. 393
Hu, M.Q.S. 385
Hu, Q.F. 595, 601
Hu, X.D. 319
Huang, H.W. 159, 215, 507, 595,
601, 643, 729, 757
Huang, M.S. 829, 849
Huang, R. 29
Huang, Z.G. 863
Idris, J. 649
Ito, H. 173
Iwata, N. 655
Jackson, P.G. 663
Jeon, S. 781
869
Jeong, K.H. 669

Jiang, J. 419
Jiang, X.H. 441
Kasper, T. 663
Kastner, R. 77, 207, 373, 689
Katebi, H. 295, 301
Kikumoto, M. 709
Kim, D.H. 555
Kim, I. 677
Kim, J. 717
Kim, S.B. 485
Kim, S.M. 399
Kim, U.Y. 555
Kokubun, T. 141
Konda, T. 173, 307, 405
Kong, X.L. 153
Koshima, A. 287
Kou, X.Q. 863
Koungelis, D.K. 785
Kulmaganbetova, Zh.K. 751
Kuzma, J. 325
Lam, S.Y. 15
Lam, T.S.K. 561
Lanquette, F. 385
Latt, K.M. 579
Le Bissonnais, H. 385
Lee, H.Y. 555
Lee, I.M. 45, 823
Lee, J.S. 555
Lee, S.D. 669, 677
Lee, S.P. 555
Lee, W. 567
Lee, Y.J. 485, 683
Leung, L.P.P. 471
Li, J. 499, 817
Li, J.P. 573, 697
Li, L. 147
Li, X.X. 165, 791
Li, Y. 689
Li, Y.Q. 153
Li, Z.H. 159
Li, Z.X. 331, 725
Liang, F.Y. 697, 849
Liang, Q.H. 455
Lin, Y.L. 165, 791
Liu, D.P. 413, 587
Liu, G.B. 227, 413, 419, 433,
587, 805
Liu, J.H. 465
Liu, S.T. 427

Liu, T. 227, 433
Liu, X. 491
Lopes, R. 287
Lu, T.D. 223
Lu, T.K. 531, 799
Lu, Z.P. 805
Lftenegger, R. 193
Ma, X.F. 307, 477
Ma, Z.Z. 477, 587
Mair, R.J. 635, 703
Makhanova, A.A. 751
Marcu, A. 187
Marlinge, J. 385
Marshall, A.M. 703
Matsumoto, A. 761
McNamara, A.M. 735
Mei, Y.B. 441
Meissner, S. 459
Michael, J. 459
Mitew-Czajewska, M. 201
Nagaya, J. 307
Nakahara, E. 709
Nakai, T. 655, 709
Negro, A. 811
Ng, C.C. 447, 579
Ng, C.W.W. 419, 433
Ng, N.W.H. 471
Niinomi, M. 655
Osborne, N.H. 447, 579
Osman, A.S. 15
Ota, H. 173, 405
Pang, P.L.R. 109, 567
Phienwej, N. 181
Pi, A.R. 319
Pieroni, M.R. 287
Popa, H. 187
Qiao, H.C. 441
Qiu, D.W. 455
Quick, H. 459
Ravaglia, A. 287
Roberts, K.J. 567
Rudi, J. 579
Ryckaert, J. 385
Rysbaeva, G.P. 751
Sabetamal, H. 295, 301
Salgaraeva, G.I. 751
Sanders, M.P.M. 261, 281
Scarpelli, G. 627
Scharinger, F. 193
Schweiger, H.F. 193
Sfriso, A. 121
Sfriso, A.O. 335
Shahin, H.M. 655, 709
Shi, Z. 817
Shin, J.H. 823
Shin, Y.J. 823
Sieminska-Lewandowska, A. 201
Soga, K. 281, 635
Song, K.I. 717
Song, T.T. 343
Song, X.Y. 829
Sozio, L.E. 837
Sugimoto, M. 761
Tabata, Y. 709
Talmon, A.M. 3, 349, 357
Tan, G.H. 579
Tang, Y.Q. 769, 843
Taylor, R.N. 627, 735
Teng, L. 313
Thiebault, H. 385
Tu, M. 551
van Tol, A.F. 261, 281
Verdel, T. 649
Viel, G. 385
Vogt, N. 275
Wang, J. 799
Wang, J.Q. 235
Wang, K.S. 331, 725
Wang, R. 413, 587
Wang, R.L. 465, 573
Wang, X.M. 729
Wang, Z.Q. 857
Wang, Z.W. 427
Wong, K.K.W. 471
870
Xia, C.C. 537

Xie, K.H. 153
Xie, X.Y. 729
Xu, Q.W. 477
Xu, S.F. 223
Xu, Y. 525
Yan, J.Y. 313, 573
Yan, Y.R. 595
Yanagawa, T. 173, 405
Yang, H.Y. 399
Yang, J.W. 669
Yang, L. 513
Yang, M. 147
Yang, S.L. 455
Yao, C.P. 601
Yao, G.S. 697
Yao, H. 607
Yao, J. 735
Ye, B. 99
Ye, G.L. 99, 307
Yoo, C. 485
Yoo, C.S. 683
Yoon, S.G. 399
You, G.M. 743
You, K. 781
Yuan, Y. 491
Zghondi, J. 207
Zhang, C.R. 849
Zhang, D.M. 215, 757
Zhang, F. 655, 709
Zhang, H. 141, 223
Zhang, J. 227
Zhang, L.M. 857
Zhang, M.X. 551
Zhang, Q.H. 269
Zhang, X. 769, 843
Zhang, Z.X. 689
Zhao, H.L. 491
Zhong, X.C. 863
Zhou, H.B. 607
Zhou, J. 153, 235
Zhou, K.Q. 455
Zhou, S.H. 343, 381
Zhu, H.H. 791
Zhu, J. 817
Zhu, W. 863
Zhu, Y.M. 441
This volume comprises a collection of four special lectures, six general

reports and 112 papers presented at the Sixth International Symposium of
Geotechnical Aspects of Underground Construction in Soft Ground (IS-Shanghai) held
between 10 and 12 April 2008 in Shanghai, China.
The Symposium was organised by Tongji University and supported by China
Civil Engineering Society, Chinese Society for Chinese Society for Rock
Mechanics and Engineering, Geotechnical Division, the Hong Kong
Institution of Engineers, Hong Kong Geotechnical Society, Hong Kong
University of Science and Technology, Science and Technology Commission
of Shanghai Municipality, Shanghai Changjiang Tunnels and Bridge
Development Co. Ltd and Shanghai Society of Civil Engineering. This was
the most recent symposium in a series of symposia starting in New Delhi,
India (1994), followed by symposia in London, UK (1996), Tokyo, Japan
(1999), Toulouse, France (2002) and Amsterdam, the Netherlands (2005).
The four invited special lectures from A. Bezuijen, Huang Rong, M.D. Bolton
and I.K. Lee are about Processes around a TBM, Overview of Shanghai
Yangtze river tunnel project, Supporting excavations in clay - from analysis
to decision-making and Underground construction in decomposed residual
soils, respectively.
The six general reports from Richard Finno, Tadashi Hashimoto, Alejo Sfriso,
C.T. Chin, Richard Pang and Richard Kastner, cover the symposium themes:
1. Analysis and numerical modelling of deep excavations
2. Construction method, ground treatment, and conditioning for tunnelling
3. Case histories
4. Safety issues, risk analysis, hazard management and control
5. Physical and numerical modelling
6. Calculation, design methods, and predictive tools
This volume provides a valuable source of information on the the state-ofthe-art in geotechnical engineering associated with the design, construction
and monitoring of tunnels and excavations in soft ground, and will be of
interest to academics and professionals involved in these areas.
an informa business

Geotechnical Aspects of Underground Construction in Soft Ground - C. NG., Et Al., (CRC, 2009) BBS

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Geotechnical Aspects of Underground Construction in Soft Ground - C. NG., Et Al., (CRC, 2009) BBS

Uploaded by

Copyright:

Available Formats

PROCEEDINGS OF THE 6TH INTERNATIONAL SYMPOSIUM (IS-SHANGHAI 2008),

H.W. Huang & G.B. Liu

ISBN: 978-0-415-48475-6 (Hardback)

Supporting excavations in clay from analysis to decision-making

Overview of Shanghai Yangtze River Tunnel Project

Underground construction in decomposed residual soils

Calculation and design methods, and predictive tools

Analysis and numerical modeling of deep excavations

Construction method, ground treatment, and conditioning for tunneling

Physical and numerical modelling

Theme 1: Analysis and numerical modeling of deep excavations

Three-dimensional finite element analysis of diaphragm walls for top-down construction

Numerical evaluation of dewatering effect on deep excavation in soft clay

Analysis of the factors influencing foundation pit deformations

Construction monitoring and numerical simulation of an excavation with SMW

A simplified spatial methodology of earth pressure against retaining piles of pile-row

Consideration of design method for braced excavation based on monitoring results

Ground movements in station excavations of Bangkok first MRT

Numerical modelling and experimental measurements for a retaining wall of

The effect of deep excavation on surrounding ground and nearby structures

Multi-criteria procedure for the back-analysis of multi-supported retaining walls

Study on deformation laws under the construction of semi-reverse method

Theme 2: Construction method, ground treatment, and conditioning for tunnelling

The influence of flow around a TBM machine

Mechanisms that determine between fracture and compaction grouting in sand

Research of non-motor vehicle-rail transit-tube interchanging transport system pattern

Shotcrete excavations for the Munich subway Comparison of different methods

The effects of sample dimension and gradation on shear strength parameters of

Experimental study on compressibility behavior of foamed sandy soil

The double-o-tube shield tunnel in Shanghai soil

The influence of engineering-geological conditions on construction of the

Critical ventilation velocity in large cross-section road tunnel fire

Metro tunnels in Buenos Aires: Design and construction procedures 19982007

Study on the earth pressure distribution of excavation chamber in EPB tunneling

Backfill grouting research at Groene Hart Tunnel

Longitudinal tube bending due to grout pressures

Theme 3: Case histories

Shield tunneling beneath existing railway line in soft ground

Case history on a railway tunnel in soft rock (Morocco)

Observed behaviours of deep excavations in sand

Environmental problems of groundwater around the longest expressway tunnel in Korea

Measurements of ground deformations behind braced excavations

Research on the effect of buried channels to the differential settlement of building

Performance of a deep excavation in soft clay

Deformation monitoring during construction of subway tunnels in soft ground

The construction and field monitoring of a deep excavation in soft soils

The benefits of hybrid ground treatment in significantly reducing wall movement:

3D deformation monitoring of subway tunnel

Challenging urban tunnelling projects in soft soil conditions

Characteristics of tunneling-induced ground settlement in groundwater drawdown environment

Effect of long-term settlement on longitudinal mechanical performance of tunnel in soft soil

Theme 4: Safety issues, risk analysis harzard management and control

Risk assessment for the safe grade of deep excavation

Multi-factors durability evaluation in subway concrete structure

Research and application of road tunnel structural optimization

Floor heave behavior and control of roadway intersection in deep mine

Framework of performance-based fire protection design method for road tunnel

Prediction of surface settlements induced by shield tunneling: An ANFIS model

Performance review of a pipe jacking project in Hong Kong

Maximising the potential of strain gauges: A Singapore perspective

Risk assessment on environmental impact in Xizang Road Tunnel

Risk analysis and fuzzy comprehensive assessment on construction of shield